Modelling of laser powder bed fusion process and analysing the effective parameters on surface characteristics of Ti-6Al-4V

https://doi.org/10.1016/j.ijmecsci.2019.105299Get rights and content

Highlights

  • To have consistent results an average surface Sa was selected and measured as a target.

  • Artificial neural network was used as an accurate tool for characterizing the interaction and effect of parameters on the surface of SLM parts.

  • The effect of SLM process parameters and heat treatment on the values of average surface were modeled accurately.

Abstract

In this paper, we propose a model to predict the average surface roughness (Sa) and analyse the effect of related process parameters on laser powder bed fusion (LPBF) selective laser melting (SLM) of Ti-6Al-4 V. The additive manufacturing (AM) process has various independent parameters that affect the quality of the produced parts and is complex to analyse. Although the process parameters can be selected separately in LPBF, they do however affect each other. Therefore, large adjustments of process parameters is not possible due to the negative effect they have on each other which can lead to problems such as cracks, balling, unmelted powders, porosity, and distortion. A range of process parameters using Taguchi L25 design of experiment (DOE) with five repetitions for each sample has been selected. Then, an artificial neural network (ANN) is applied to the model to predict the value of (arithmetical mean height)/(average surface roughness) (Sa). The selected processing parameters are laser power, scan speed, hatch spacing, laser pattern increment angle, and heat treatment (HT) condition. The present work revolves around ANN modeling and using a wide parameter range and a large number of test samples under ASTM standards as well as adding HT to the DOE to analyse the simultaneous effect of HT and changing process parameters on surface characteristics. A large and precise data set with high generality and reliability obtained by 3750 profilometries on 125 samples. The contribution of this paper is using ANN as an accurate tool in surface modeling and characterizing the effective parameters on the surface of LPBF parts. The existence phenomena and governing factors were explained by introducing new parametric mechanisms in rheology of melting pool. In AM of metals, the variation of average roughness in overlap of hatches can be 5–7 times higher than the centre of the track. Therefore, Sa was selected to have consistency in the measured roughness values. Results showed heat treatment above beta phase transus leads to a local flow of material at the surface causing an increase of Sa. The ranking of influential factors on Sa from the highest to the lowest was found to be: heat treatment > laser power > scan pattern angle > hatch space > scan speed.

Introduction

Ti-6Al-4V is one of the most common titanium alloys with particular properties such as high specific strength, excellent corrosion resistance and good fatigue resistance. These properties characterize it as one of the best materials for lightweight and high strength components used in aerospace and biomedical applications. Surface quality and dimensional accuracy are among the critical factors in manufacturing for such applications, especially for parts which are assembled in multi-component systems. Selective laser melting (SLM) is additive manufacturing (AM) technology, sub-classified as in laser powder bed fusion (LPBF), that generates poor surface quality, often making it necessary to carry out further post-processing and metal cutting [1], [2], [3].

During LPBF, a few defects may occur with negative effects on the process and surface quality of the produced parts. Porosity is a well-known problem in AM of metals [4], generally referring to pores created during laser material processing which are mostly classified into two types: keyhole and balling. Keyholes generally occur when a huge amount of energy is focused on a small area resulting in the formation of a melt pool which is narrow and deep. Gasses trapped in this region form a classical keyhole shaped void [5], [6]. Semak and Matsunawa [7] found that balling defects in laser material processing were mainly created due to the high flow of fluids in the melt pool that is affected by temperature and surface tension. Teng et al. [8] reviewed the occurrence of defects in the laser melting process and how it can be modelled. A literature survey of thermal modelling in metal laser sintering shows the development of various models by combining material thermal nonlinearity, latent heat, laser heat source distribution and (interaction between) the laser beam and powder bed, leading to the generation of various surface qualities [9].

Powder characteristics and process parameters are key factors affecting surface quality during the LPBF process. The connections between the foremost LPBF parameters (laser control thickness, scanning speed, and layer thickness), properties of the powder and geometrical attributes of single tracks have been investigated by Yadroitsev [10]. This investigation shows smaller particles can fill small voids between larger powder particles and improve density and surface quality. Finer particles provide a larger surface area and absorb more laser energy, thus generating bigger melting pools and improving surface quality. The surface quality of SLM CoCrMo parts has been shown to be affected by process parameters. Single layer analyses on surface quality were carried out to identify optimum input parameters. For the manufacture of parts with a smooth surface, the scan space should be kept low. Also during the melting process, surface tension affects the roughness [11], [12]. The particle size and distribution was also shown to change the surface quality of LPBF parts. Processing of powders with coarser particles results in higher porosity and a rougher surface [13], [14], [15]. By increasing the scan speed, balling was observed due to low wettability, reduced contact between substrate and melt, and Rayleigh instability. Large overlap occurs with low hatch spacing, which increases the roughness of the surface [16].Thijs et al. [17] assumed that the formation of pores in a Ti and Al-based alloy is associated with powder deposited around the melt pool within a layer and an accumulation of the surface roughness across the layers and also the formation of keyholes. These phenomena were therefore thought to affect surface roughness and decrease the value of Ra and surface quality of produced parts [18]. The size of droplets in the balling region tended to increase with increasing laser power and decreasing scan speeds and a large softening area was identified which changed the surface quality of 24-carat gold. In this investigation, the vast majority of the porosity was observed to be between two subsequent layers [19]. Colignano et al. [20] showed the choice of parameters of the STL file changes the accuracy and dimensional deviations without creating too many triangles. This can improve the design and increase the speed of the process.

In order to observe the contour profile of the melt pool with respect to the scan types, LPBF parts were fabricated with various hatch spacing while keeping other process parameters constant. Controlling and optimization of the process parameters and modification of the initial design are key factors in the tailoring the surface quality of AM parts. For instance, the scan speed is directly proportional to a physical density where it reaches a maximum at higher scan speeds for the same energy density. However, increasing the scan speed reduces the surface quality of the parts. Therefore, it is necessary to optimize the process parameters in order to achieve a high-density material with good surface quality. However, the optimization of process parameters is not an easy task as most of the materials exhibit different conductivity and reflectivity [16], [21], [22], [23], [24], [25]. Savalani et al. [26] showed that a preheating process improves the surface of magnesium parts used in bio-implant applications. This is also shown for Ti-6Al-4V parts made by SLM where the numerical response surface (RSM) revealed that surface roughness and microstructure are greatly associated with process parameters and preheating. However, preheating has a lower impact on microstructure compared to scan speed [27]. Using larger powers slightly decreases the surface roughness by reducing the formation of balling and increasing wettability of the melt pool [28], [29]. A fractional design approach [30] has been proposed to improve the surface quality when a complex objective function exists and by optimizing process parameters in SLM of steel 17-4 PH to improve surface quality. Surface properties of the parts fabricated by LPBF highly depend on each single laser-liquefied track and every single layer, and additionally the quality of the interaction between them. The most common ways to deal with the issue of dimensional accuracy and surface quality is by experimental models, by means of design of experiment, ANN and analyse of vriance (ANOVA) and different optimization methods [31]. To model and predict the effect of process parameters on part quality various approaches such AI-based methods are available. Our previous research showed that ANN exhibit the potential to predict and analyze the effect of process parameters on density with high accuracy [32]. To optimize the density, shape, dimensional accuracy and surface quality of LPBF parts, utilization of laser surface re-melting (LSR) has been suggested [33], [34]. Post-processing such as metal cutting, ball-pinning and laser or electron beam remelting are suggested to solve the mechanical and surface problems of PBF parts [12], [35], [36], [37], [38].

All discussed methods to analyse the LPBF process are highly reliant on the accuracy of transfer functions that is related to the number of test sets. Using a small number of test specimens due to the cost of each experiment results in generating low accuracy of the presented models. Also, in the reviewed investigations the average roughness was reported which can vary significantly between two subsequent measurements and is associated with the place of measurement (e.g. in overlap regions or at the center of the hatches).

Therefore, in this research to obtain more accurate results, an artificial neural network (ANN) is used to model, predict and analyse the effective parameters in the LPBF process. In this work, based on Taguchi L25, samples were printed with different process and post-process parameters including laser power (LP), scan speed (SS), hatch space (HS), pattern angle (PA) and heat treatment (HT). Then surfaces were investigated using non-contact propfilometry method and the results were used in an ANN method to develop a predictive model for process parameters. Analysing the interaction of each parameter revealed the ranking of influential factors on average surface. The existence phenomena are drawn by introducing the rheological mechanisms of the melting pool.

Section snippets

Powder material and LPBF operation

An SLM Solutions 125HL LPBF machine with maximum laser power 200 W equipped with a YLR-Fiber-Laser and minimum spot size 5 μm was used. Fig. 1(A) illustrates spherical Ti-6Al-4 V powder used for this research. In order to achieve samples with high density, a meander scanning pattern with an incremental angle change in every layer has been used to print dog-bone samples based on ASTM E8-E8M standard for further works. Fig. 1 (B, C) shows the CAM process, (D) shows the powder particle size

Surface roughness measurement

An optical profilometer (Alicona Infinite Focus) equipped with 5–100× objective lenses was used for scanning the printed samples and surface roughness measurement to investigate the surface characteristics. For all conditions, five different areas of the topmost surface, each with 10 × 5 mm size and containing 800 points, were scanned according to the standard requirements. The average roughness of the as-built samples generally shows significant fluctuations (sometimes as high as 20 fold) due

Analysing the obtained data from DOE

To validate the accuracy and the trend of the obtained data, signal-to-noise (SN) ratio needs to be calculated based on the Taguchi SN relationship. Generally, signal-to-noise shows the ratio of signal or the obtained values to the noise (errors). This criterion shows the accuracy and performance of the measurement in each experiment. In the production of LPBF parts, one of the problems for as-built samples is low surface quality so improving the roughness is a goal. Therefore, the criteria

Average contribution of input nodes on outputs using ANN and analyse of variance (ANOVA)

To determine the most effective process parameters on the average surface an interrogator analysis using the proposed ANN was carried out. Moreover, to prove this procedure ANOVA analyses were performed and the most effective parameters on the alternation of the average surface were determined. The average contribution of input nodes on outputs was obtained from ANN analysis and ANOVA. In ANOVA, based on standard procedures, the probability value of 0.05 was selected to determine the effective

Conclusions

This research investigates the key contribution parameters influencing average surface roughness and shows a numerical model to predict Sa and provides practical information to explain the behavior of process parameters on average roughness and enhance the surface quality of LPBF parts. This can therefore help to minimize post-treatment such as machining and polishing. The obtained data can be used to adjust the process parameters to produce higher surface quality, especially when using hybrid

Declaration of Competing Interest

We confirm that we don't have a conflict of interest with other scientists in the field.

References (71)

  • G. Tapia et al.

    Prediction of porosity in metal-based additive manufacturing using spatial Gaussian process models

    Add Manuf

    (2016)
  • Z. Zhan et al.

    Development of a novel fatigue damage model with AM effects for life prediction of commonly-used alloys in aerospace

    Int J Mech Sci

    (2019)
  • A.M. Khorasani

    Investigation on the effect of cutting fluid pressure on surface quality measurement in high speed thread milling of brass alloy (C3600) and aluminium alloy (5083)

    Measurement

    (2016)
  • B. Vrancken

    Heat treatment of Ti6Al4V produced by selective laser melting: microstructure and mechanical properties

    J Alloys Compd

    (2012)
  • L. Murr

    Microstructure and mechanical behavior of Ti–6Al–4V produced by rapid-layer manufacturing, for biomedical applications

    J Mech Behav Biomed Mater

    (2009)
  • H. Attar

    Manufacture by selective laser melting and mechanical behavior of commercially pure titanium

    Mater Sci Eng

    (2014)
  • C. Weingarten

    Formation and reduction of hydrogen porosity during selective laser melting of AlSi10Mg

    J Mater Process Technol

    (2015)
  • D. Gu

    Densification behavior, microstructure evolution, and wear performance of selective laser melting processed commercially pure titanium

    Acta Mater

    (2012)
  • M. Jovanović

    The effect of annealing temperatures and cooling rates on microstructure and mechanical properties of investment cast Ti–6Al–4V alloy

    Mater Des

    (2006)
  • X. Li

    Effect of substrate temperature on the interface bond between support and substrate during selective laser melting of Al–Ni–Y–Co–La metallic glass

    Mater Des

    (2015)
  • B. Baufeld et al.

    Additive manufacturing of Ti–6Al–4V components by shaped metal deposition: microstructure and mechanical properties

    Mater Des

    (2010)
  • B. Gorny

    In situ characterization of the deformation and failure behavior of non-stochastic porous structures processed by selective laser melting

    Mater Sci Eng

    (2011)
  • S.L. Sing et al.

    Selective laser melting of titanium alloy with 50 wt% tantalum: effect of laser process parameters on part quality

    Int J Refract Met Hard Mater

    (2018)
  • W.E. King

    Observation of keyhole-mode laser melting in laser powder-bed fusion additive manufacturing

    J Mater Process Technol

    (2014)
  • S.M. Thompson

    Additive manufacturing of heat exchangers: a case study on a multi-layered Ti–6Al–4V oscillating heat pipe

    Add Manuf

    (2015)
  • M. Shukla

    Effect of laser power and powder flow rate on properties of laser metal deposited Ti6Al4V. World Academy of Science, Engineering and Technology

    Int J Mech Aerosp Ind Mechatron Manuf Eng

    (2012)
  • A. Mahyar Khorasani

    Characterizing the effect of cutting condition, tool path, and heat treatment on cutting forces of selective laser melting spherical component in five-axis milling

    J Manuf Sci Eng

    (2018)
  • I. Gibson et al.
    (2010)
  • D.W. T. Nguyen et al.

    High speed fusion weld bead defects

    Sci Technol Weld Join

    (2013)
  • G.C. S. Li et al.

    Relationship between spatter formationand dynamic molten pool during high-power deep-penetration laserwelding

    Appl Surf Sci

    (2014)
  • A.M. V. Semak

    The role of recoil pressure in energy balanceduring laser materials processing

    J Phys D

    (1997)
  • D.P. C. Teng et al.

    A review of defect modeling in laser material processing

    Add Manuf

    (2016)
  • D.P.a.B.S. K. Zeng

    A review of thermal analysis methods in laser sintering and selective laser melting

    (2017)
  • I. Yadroitsev

    Factor analysis of selective laser melting process parameters and geometrical characteristics of synthesized single tracks

    Rapid Prototyp J

    (2012)
  • K.M.a.J.C. Y. Pupo

    Influence of process parameters on surface quality of CoCrMo produced by selective laser melting

    Int J Adv Manuf Technol

    (2015)
  • Cited by (0)

    View full text