Effect of processing parameters during the laser beam melting of Inconel 738: Comparison between simulated and experimental melt pool shape

https://doi.org/10.1016/j.jmatprotec.2020.116897Get rights and content

Highlights

  • A novel calibration method is proposed to calculate the absorption coefficient.

  • The numerical method is able to predict the melt pool shape on a wide process window.

  • The transition from low keyhole melt pool morphology to deep keyhole is correctly described.

  • Metal ejections during laser-matter interaction need to be considered in the simulation.

Abstract

Numerical simulation is a powerful tool to understand the link between processing parameters and solidification conditions during the laser beam melting (LBM) process. To be able to use this tool for microstructural control, numerical models need to be validated on a large set of experimental conditions, to ensure that the model describes the predominant physical phenomena. In this study, an experimental set of twenty tracks was produced in an Inconel 738 alloy, with a wide range of energy input and scanning speed. Experimental melt pool shapes were compared to the predictions of a multiphysics numerical model. In this model, the powder bed is considered as a continuum. The laser source is modeled with a Beer-Lambert absorption law, and surface tension, Marangoni force and recoil pressure are the driving forces for melt pool dynamics. This kind of model offers an efficient computational time, but requires a calibration of the absorption coefficient and a representative description of laser-matter interaction. In order to represent correctly heat and mass transfer during laser-matter interaction, the model needs to account for the loss of matter caused by the ejection of powder particles and spatters. A novel calibration method was proposed to calculate the absorption coefficient. This method uses the experimental cross sections of the melt pools and a simplified analytical expression of energy balance. The use of this calibration method enabled a good agreement between experiments and calculations on a large process window. The values obtained by the calibrations resulted in a phenomenological expression of absorptivity coefficient with process parameters. Based on this expression, a comparison was made with another numerical model from literature using a time-consuming ray-tracing method in order to calculate the absorptivity coefficient. Similar results have been obtained, demonstrating the potential of the proposed approach to predict the melt pool shape and thus better understand the combined effect of laser-matter interaction and solidification in LBM process.

Introduction

Inconel 738 (IN738 LC) is a Nickel-based superalloy of great interest in aerospace industry, because of its excellent mechanical properties in high temperature environment, such as in aircraft engines, where parts undergo service temperatures higher than 900 °C. As a consequence, the mastering of additive manufacturing (AM) of Inconel 738 parts by laser beam melting (LBM) is a real challenge. Indeed, the high sensitivity of Inconel 738 to solidification cracking as known from the welding community makes its manufacturing by LBM process critical. This alloy is difficult to weld, and hence it is also difficult to process additively. The challenge is then to identify appropriate LBM "process window" (set of process parameters) adapted to a fabrication without any defect.

In the context of AM-LBM, solidification cracking of Inconel 738 has been experimentally investigated in several studies, some of which focus on the influence of processing parameters. Cloots et al. (2016) have shown that the number of microcracks tends to decrease when the scanning velocity is increased while the beam power is maintained constant. Grange et al. (2020) have shown that the cracking is minimal when the material is processed with small melt pools with a strong overlap. They mentioned four contributing factors for solidification cracking: the extent of the mushy zone, the intensity of stresses in the mushy zone, a positive role of a fine grain structure and a positive effect of material remelting to avoid crack propagation.

Beside experiments, numerical simulation is a powerful tool to understand and control materials processing. The development of numerical simulation, assessed and validated by experimental observations and measurements, allows engineers to develop strategies to identify adequate process windows. However, it should be observed in LBM context that a very complex physics is at stake. Therefore, addressing directly the prediction of solidification cracking through numerical simulation might be a too ambitious objective, as this would require predictive models in laser/metal interaction, fluid flow, solidification, formation of microstructure, and thermo-mechanics in the semi-solid state. It appears then that a more progressive approach in numerical simulation development should be preferred, with a first assigned objective: a thermo-fluid numerical model capable of calculating a reliable description of LBM solidification conditions. Indeed, predicting the melt pool shape, the extension of the mushy zone, as well as the temperature gradients and cooling rates locally in the vicinity of the melt pool and in the mushy zone, is an obvious prerequisite before addressing thermo-mechanics. This is precisely the objective of the present paper: the evaluation of a multiphysics thermo-fluid simulation model by reference to experimental measurements, for LBM of Inconel 738.

Numerous experimental studies highlighted the complexity and the multiplicity of the physical phenomena at stake and show which ones are essential to consider in numerical models. Experimentally, Yadroitsev and Smurov (2010) demonstrated the influence of process parameters such as scanning speed and laser power on single track formation for different alloys, including IN625. They demonstrated that melt pool penetration into the substrate is required to stabilize the track building and avoid track irregularities of balling. Furthermore, the tracks shape and dimensions are largely influenced by surface tension and Marangoni forces. Bidare et al. (2018) used fast camera equipment to develop observations of tracks development and vaporization stage during LBM process for stainless steel. They demonstrated that this latter phenomenon highly influences the melt pool and particles dynamics. Wang et al. (2017) suggested that the observed liquid spattering is a consequence of the recoil pressure induced by vaporization combined with Marangoni effect using CoCr powder. Furthermore, as shown by Matthews et al. (2016), the recoil pressure is also partly responsible for the denudation on track sides. From this short literature review, it appears that a predictive numerical model should at least take into account the following physical phenomena: heat transfer, fluid flow, surface tension including Marangoni effect, and laser/matter interaction including vaporization effect.

The numerical simulation community has worked for several years to develop computational codes matching the previous requirements. Complex models have been developed to predict the melt pool and final track shape. Most of them consider the scale of powder particles, with an explicit description of every particle of the powder bed. Khairallah et al. (2016) presented an arbitrary Lagrangian–Eulerian (ALE) multiphysics code to simulate laser-matter interaction and fluid flow with an application to Ti-6Al-4 V. The Marangoni effect and the recoil pressure are considered in order to predict denudation during the heating stage. Martin et al. (2019) used this model to provide better understanding of the occurrence of porosities during the transition from a track to another one. Bayat et al. (2019) developed a similar multiphysics model and applied it to Ti-6Al-4 V. They were able to get an accurate prediction of the melted zone dimensions compared to experimental observations. Aggarwal et al. (2019) used a similar model on 316 L stainless steel. They used the simulation to understand the influence of the distance between the powder bed position and the focal plane of the laser beam on the resulting melting mode (conduction or keyhole). However, being based on an explicit discretization of the powder bed particles, and possibly ray-tracing for laser/metal interaction, all these approaches are still excessively time consuming to model track evolution. This limits length of simulation domains even more when the formation of multiple tracks and layers is considered, as in effective AM. In view of an efficient search for process windows, a reasonable computational time is required as well as the development of a reliable model. Consequently, the simulation model previously proposed in Queva et al. (2020) is considered in the present study. In this approach the powder bed is modelled as a continuum and a Beer-Lambert absorption law is assumed to consider the progressive absorption of laser energy in matter, which generates lower computational time than previous ones. Queva et al. (2020) reported that the CPU time required to simulate a single track is approximately 3.5 times smaller than with other approaches. However, as a counterpart, the laser absorptivity is a variable parameter which has to be calibrated. In this article, a novel method was proposed to calibrate the absorptivity coefficient, with a combination of semi-analytical reasoning and numerical simulations in order to investigate the effect of processing parameters on experimental melt pool cross sections.

The study reported here was structured as follows:

  • The evolution of the melt pool shape of IN738 was studied, for a wide range of LBM processing parameters. A set of twenty experiments was carried out under different laser power and scanning speed to provide a process window delimiting parameters where the track is stable from those associated to keyhole or capillary instabilities. A study of mass transfer during the interaction between the laser beam and the powder bed was conducted, based on optical profilometry measurements, and revealed the necessity to account for metal ejections. This is reported in Section 2.

  • Experimental observations were compared with the predictions of the numerical model developed by Queva et al. (2020). A novel calibration method for the absorptivity coefficient of the numerical simulation was proposed, based on the experimental cross sections of the melt pool and a simplified analytical expression of energy conservation. This is reported in Section 3.

  • After calibration, numerical simulation results and experimental observations were compared for the whole process window, demonstrating the capabilities of the present numerical approach to provide an efficient prediction of the melted zone dimensions, in a wide range of process parameters, and with an efficient computational time. The transition from very low keyhole melt pool morphology to deep keyhole is presented. An investigation of the absorptivity coefficient evolution with processing parameters is proposed and compared with in-situ measurements from literature, showing a good agreement. This is reported in Section 4.

Section snippets

Experimental protocol

All the experiments reported here were undertaken with a gas atomized powder of Inconel 738 LC, and a Concept Laser M2 machine. The particle size distribution is Gaussian, with diameter percentiles D10=17 μm, D50=28 μm and D90=45 μm. The substrate is a cylinder of 30 mm diameter and 10 mm height, previously fabricated with the same printer and the same material. The upper surface was then polished with 1200-grit paper and sandblasted to prevent the powder from sliding on the substrate. A

Numerical model

A continuous-mesoscale finite element model thereafter applied to simulate the development of the set of single tracks was previously proposed for the investigation of melt pool development on metallic alloys during LBM process (Queva et al., 2020). This model relies on a level set (LS) formulation of conservation equations. Basically, the simulation domain is shared into two parts associated with the metallic material and the protective atmosphere. The metallic material includes the substrate,

Melt pool dimensions

Even if the present model is able to predict correctly the melt pool dimension, it tends to slightly underestimate the apparent height Happ. One of the possible explanations is that some particles are drawn from both sides of the track toward the melt pool, as observed by Bidare et al. (2018). Here, because the powder bed is modelled as a continuum, this effect is not considered in the simulation, leading to an underestimation of Happ. Particles dynamics with movements towards the melt pool and

Conclusion

In this study, a set of tracks were fabricated in an Inconel 738 LC nickel-base superalloy using LBM process with a large set of processing parameters. The dimensions of tracks were precisely measured and cross sections of melted domains observed. Processing parameters with a linear incident energy between 0.22 and 0.5J.mm-1 lead to tracks with stable dimensions and a penetration ratio HRZ/Happ between 1 and 3. The use of analytical models allowed to describe the melt pool cross section, but

CRediT authorship contribution statement

D. Grange: Conceptualization, Methodology, Investigation, Formal analysis, Writing - original draft. A. Queva: Methodology, Software, Investigation, Writing - original draft. G. Guillemot: Methodology, Validation, Writing - review & editing. M. Bellet: Methodology, Writing - review & editing, Supervision. J.-D. Bartout: Methodology, Resources. C. Colin: Methodology, Validation, Supervision.

Declaration of Competing Interest

The authors report no declarations of interest.

Acknowledgements

Safran Tech (Châteaufort, France) funded this study. The authors would like to thank Clara Moriconi and Bruno Macquaire for their constructive remarks.

References (26)

Cited by (33)

View all citing articles on Scopus
View full text