Melt Pool Depth Research Articles

Cross-Sectional Melt Pool Geometry of Laser Scanned Tracks and Pads on Nickel Alloy 718 for the 2022 Additive Manufacturing Benchmark Challenges

The Additive Manufacturing Benchmark Series (AM Bench) is a NIST-led organization that provides a continuing series of additive manufacturing benchmark measurements, challenge problems, and conferences with the primary goal of enabling modelers to test their simulations against rigorous, highly controlled additive manufacturing benchmark measurement data. To this end, single-track (1D) and pad (2D) scans on bare plate nickel alloy 718 were completed with thermography, cross-sectional grain orientation and local chemical composition maps, and cross-sectional melt pool size measurements. The laser power, scan speed, and laser spot size were varied for single tracks, and the scan direction was varied for pads. This article focuses on the cross-sectional melt pool size measurements and presents the predictions from challenge problems. Single-track depth correlated with volumetric energy density while width did not (within the studied parameters). The melt pool size for pad scans was greater than single tracks due to heat buildup. Pad scan melt pool depth was reduced when the laser scan direction and gas flow direction were parallel. The melt pool size in pad scans showed little to no trend against position within the pads. Uncertainty budgets for cross-sectional melt pool size from optical micrographs are provided for the purpose of model validation.

Effect of process parameters on the roughness and tensile behavior of parts manufactured by the metals LPBF process

AbstractThe laser powder bed fusion process for metals (LPBF‐M) results in the development of stochastic surface features that significantly influence the interactions between parts and their surrounding environment, as well as their mechanical properties. The process parameters influence the surface quality, which is quantified by the surface roughness. Therefore, customizing the surface roughness during the build process can significantly contribute to obtaining ready‐to‐use parts, reducing the need for extensive surface posttreatments. This paper utilizes theoretical estimations of melt pool depth under iso‐linear energy density, iso‐power, and iso‐temperature manufacturing process parameter conditions. These estimations are then compared with experimental evaluations of surface roughness and tensile strength in upright‐built specimens to extract the trends in terms of the input energy versus roughness, and the input energy versus tensile behavior. The results show that iso‐energy values yield similar roughnesses due to the consistent expected melt pool depth. Moreover, an increase in melt pool depth generates higher surface roughness, while smaller melt pool dimensions result in improved roughness. Additionally, a comparison between the melt pool size and tensile test performance reveals a detrimental impact on the tensile strength for specimens estimated to have smaller melt pool depths.

A novel feature engineering approach for predicting melt pool depth during LPBF by machine learning models

Melt pool geometry is a deterministic factor affecting the characteristics of metal Additive Manufacturing (AM) components. The wide array of physical and thermal phenomena involved during the formation of the AM melt pool, along with the great variety of alloy compositions and AM methods, coupled with the clear influence of multiple process parameters, make it difficult to predict the melt pool geometry under a given set of conditions. Therefore, using Artificial Intelligence (AI) approaches such as Machine Learning (ML) is necessary for accurate predictions. Using a physics-informed feature selection strategy along with the application of atomic features for the first time, this work aims to offer accurately trained models relying on existing high-fidelity data for most common alloys in AM academia and industry, i.e., 316 L stainless steel, Ti6Al4V, and AlSi10Mg. Multiple ML algorithms were trained, and the results revealed that the average R2 and RMSE obtained by the K-fold cross-validation (K = 5) were significantly enhanced when laser and material properties, inspired by the analytical models for AM melt pool geometry, were used as the model features. Removing the excess features and applying atomic features further enhanced the accuracy of the models. As a result, R2 for the XGBoost, CatBoost, and GPR models were 0.907, 0.889, and 0.882, respectively, while the hold-out cross-validation led to 0.978, 0.976, and 0.945, respectively. Furthermore, the results showed that the XGBoost model outperforms the Rosenthal equation. This approach provides a pathway to more accurately predict the properties of metal AM components.

A Novel Physics-Based Model for Predicting Melt Pool Dimensions in Laser Powder Bed Fusion Process

Abstract This paper employed a scaling analysis to represent the processing parameters, affecting the melting process in the dimensionless numbers, identify the relationships of these dimensionless numbers, and develop semi-empirical correlations to predict the width and depth of the melt pool. To develop the correlations, Ti-6Al-4V powder was used to print 38 tracks at various processing conditions. The correlations were then fit into this experimental data using python code to find the constants of the correlations. The correlations were then used to predict the depth and width of the melt pools. It was found that the mean discrepancy between the predicted melt pool dimensions and the experiment is 7%. To evaluate the accuracy of the correlation in predicting the melt pool dimensions of the materials never used during the development of the correlations, the melt pool depth of some tracks made out of stainless steel 316L printed at various conditions was predicted using the model, and it was found that the mean discrepancy between the predicted melt pool depth and experiment is 11%.

Geometrically-informed predictive modeling of melt pool depth in laser powder bed fusion using deep MLP-CNN and metadata integration

The laser powder bed fusion (LPBF) process has the ability to manufacture intricate geometries. However, the fabrication of complex geometries using the LPBF process for industrial applications remains challenging in terms of achieving consistent microstructure and mechanical properties. Experimental investigations show that in addition to process parameters, geometrical factors such as shape and size of parts impact the thermal history, microstructure, defect structure, and thus mechanical and fatigue properties. As a result, the standardization, qualification, and certification of additively manufactured (AM-ed) parts encounters substantial obstacles posed by the obscure impact of design parameters on component quality. In this paper, an artificial intelligence framework is proposed to predict the melt pool depth of the additively manufactured parts considering the geometrical information of the parts. More specifically, a deep Multilayer Perceptron and Convolutional Neural Network (MLP-CNN) framework is designed to predict melt pool depth for any part geometries when printing metallic parts via LPBF using metadata (i.e., numerical and image data). Quite often the melt pool depth generated during the AM processes is employed as a surrogate for thermal, defect- and micro-structural signatures. In melt pool depth prediction of complex geometries in LPBF, a digital twin environment using a finite element model (FEM) is developed and validated using experimental melt pool measurements for different geometries within multi-tracks and layers. The FEM was in agreement with experimentation with an error of less than 15%. Through the process simulation, a large dataset of melt pool depths was obtained for various part geometries and different process parameters. Next, the MLP-CNN framework was established to identify the parallel impact of part design and process parameters on melt pool depths. To enhance the performance and robustness of this model, data augmentation has been implemented by rotating and transferring geometries to artificially expand the dataset. Data augmentation helps to mitigate overfitting and promote better generalization, especially in the context of limited training data. After training, the proposed model is found to give accurate melt pool prediction even for new geometries not considered during training. The fusion of CNN and MLP has led the melt pool predictions for unseen geometries with an accuracy of 95%.

Assessing powder processability and melting behavior in powder-bed fusion additive manufacturing

In this study, the effects of powder characteristics on powder bed properties and the subsequent impacts of these properties on melt solidification were investigated. To this end, eight stainless-steel powders with varying characteristics were examined. The process maps generated for these powders were also assessed, focusing on their melt-solidification behavior and their implications for buildability (i.e., product quality). The build density distribution across the process map exhibited substantial variations. This variability was primarily attributed to the distinct melting behaviors associated with the unique characteristics of each powder. Both the real powder layer thickness of the powder bed and the melt-pool depth could be estimated by performing numerical simulations. Powder characteristics, particularly particle size and particle-size distribution, had significant effects on the achieved build density. Considering these findings, we introduced a new parameter zR/D90, which is determined by the powder characteristics and real powder layer thickness. Consequently, zR/D90 emerged as a valuable parameter for predicting the buildability of the powder. Therefore, this parameter can be noted as a reliable predictor of powder buildability, thereby providing guidelines for manufacturing parts with few defects.

Laser metal deposition of titanium on stainless steel with high powder flowrate for high interfacial strength

Direct joining of titanium and stainless steel 316 L with a strong interface is very challenging due to the formation of the brittle intermetallic compounds FeTi and Fe2Ti in the intermixing zones and to the high residual stress induced by the mismatch of the thermal expansion coefficients. In this bimetallic directed energy deposition study, firstly, deposition of Ti on stainless steel was carried out using conventional process parameter regime to understand the interfacial cracking susceptibility and then a novel high powder flowrate approach is proposed for controlling the dilution and constraining the intermetallic phases forming at the interface. The influence of high temperature substrate preheating (520 °C) on the cracking susceptibility and interface strength was also investigated. The deposited Ti samples and their interfaces with the 316 L substrate were characterized with optical microscopy, scanning electron microscopy and energy dispersive X-ray spectroscopy to investigate the geometry, microstructures and chemical compositions in relation to the cracks. The high powder flowrate deposition of Ti on stainless steel 316 L results in an extremely thin dilution region (~ 10 μm melt pool depth in the substrate) restricting the formation of the intermetallic phases and cracks. The ultimate shear strength of the interfaces of the crack free sample was measured from cuboid deposits and the highest measured strength is 381 ± 24 MPa, exceeding the weaker base material pure Ti. The high interfacial strength for high powder flowrate deposition is due to the substantial attenuation and shadowing of the laser beam by the in-flight powder stream as demonstrated by the high-speed imaging resulting in an extremely small dilution region.

Alternate deposition and remelting microdroplets via single laser for printing low-defect and high-performance metal micropillars

Laser-induced forward transfer (LIFT) has emerged as a versatile technique for printing high-resolution metal microstructures. However, a common drawback of this method is the inadequate coalescence of the deposited metal microdroplets, which results in inferior electrical and mechanical properties. This paper proposes a novel approach for fabricating high-performance metal micropillars using a single-pulsed laser to alternately deposit and remelt metal microdroplets. Specifically, an ultraviolet nanosecond laser was used to induce the deposition of copper microdroplets, forming a patterned powder bed with high resolution. Subsequently, a laser pulse train was applied to fuse the patterned powder bed. The results showed that voids and microdroplet delamination were eliminated in the printed copper micropillars, whose yield strength and elastic modulus increased threefold, approaching 63% of those of the bulk metal. The remelting behavior of the deposited microdroplets was elucidated by modelling and analysing the thermal accumulation effects of a laser pulse train. A remelting map was proposed, including the non-melting, remelting, and vaporizing regimes. According to the depth of melt pool, the evolutions of morphology and microstructure in the depositing and remelting process were elucidated. Hence, this study advances the LIFT process for fabricating high-performance metal microstructures.

Impact of Boundary Parameters Accuracy on Modeling of Directed Energy Deposition Thermal Field

Within the large Additive Manufacturing (AM) process family, Directed Energy Deposition (DED) can be used to create low-cost prototypes and coatings, or to repair cracks. In the case of M4 HSS (High Speed Steel), a reliable computed temperature field during DED process allows the optimization of the substrate preheating temperature value and other process parameters. Such optimization is required to avoid failure during the process, as well as high residual stresses. If 3D DED simulations provide accurate thermal fields, they also induce huge computation time, which motivates simplifications. This article uses a 2D Finite Element (FE) model that decreases the computation cost through dividing the CPU time by around 100 in our studied case, but it needs some calibrations. As described, the identification of a correct data set solely based on local temperature measurements can lead to various sets of parameters with variations of up to 100%. In this study, the melt pool depth was used as an additional experimental measurement to identify the input data set, and a sensitivity analysis was conducted to estimate the impact of each identified parameter on the cooling rate and the melt pool dimension.

Computational Modeling of Directed Energy Deposition of Titanium Aluminide Plate Geometries

In this work, transient thermomechanical analysis of the laser-based Directed Energy Deposition (DED) process was carried out to predict melt pool dimensions, thermal cycles and residual stress during Titanium Aluminide (TiAl) plate deposition. The predicted melt pool length, width, and depth increased as the laser power was increased from 250 W to 350 W at a constant travel speed of 15 mm/s. The melt pool length increased from 0.425 mm to 0.55 mm, while the depth increased from 0.1 to 0.29 mm, and the width increased from 0.412 to 0.5 mm. The plate's edges (tracks 1 and 6) were found to experience a slightly higher temperature than the middle tracks, which may contribute to higher thermal gradients at the edges. The highest Principal stress of 140 MPa was found in track 1. The magnitude of Principal stresses diminished as track deposition progressed from track 2 to 6 due to the variation of the thermal gradient along the tracks.

Predicting meltpool depth and primary dendritic arm spacing in laser powder bed fusion additive manufacturing using physics-based machine learning

The long-term goal of this work is to predict and control the microstructure evolution in metal additive manufacturing processes. In pursuit of this goal, we developed and applied an approach which combines physics-based thermal modeling with machine learning to predict two important microstructure-related characteristics, namely, the meltpool depth and primary dendritic arm spacing in Nickel Alloy 718 parts made using the laser powder bed fusion (LPBF) process. Microstructure characteristics are critical determinants of functional physical properties, e.g., yield strength and fatigue life. Currently, the microstructure of LPBF parts is optimized through a cumbersome build-and-characterize empirical approach. Rapid and accurate models for predicting microstructure evolution are therefore valuable to reduce process development time and achieve consistent properties. However, owing to their computational complexity, existing physics-based models for predicting the microstructure evolution are limited to a few layers, and are challenging to scale to practical parts. This paper addresses the aforementioned research gap via a novel physics and data integrated modeling approach. The approach consists of two steps. First, a rapid, part-level computational thermal model was used to predict the temperature distribution and cooling rate in the entire part before it was printed. Second, the foregoing physics-based thermal history quantifiers were used as inputs to a simple machine learning model (support vector machine) trained to predict the meltpool depth and primary dendritic arm spacing based on empirical materials characterization data. As an example of its efficacy, when tested on a separate set of samples from a different build, the approach predicted the primary dendritic arm spacing with root mean squared error ≈ 110 nm. This work thus presents an avenue for future physics-based optimization and control of microstructure evolution in LPBF.

Influence of temperature and beam size on weld track shape in laser powder bed fusion of pure copper using near-infrared laser system

Additive manufacturing of copper using laser powder bed fusion enables the production of highly complex components with excellent heat and electrical conductivity. However, the processing of copper by means of near-infrared laser radiation, which is commonly used, is challenging due to its high reflectivity. Nevertheless, it has been demonstrated that high densities and electrical conductivities can be achieved using high-power laser systems. In order to process pure copper with reliable quality with different machines, it is essential to understand the conditions at which a continuous weld track is formed. For this purpose, weld tracks with varying laser power and scan speeds were welded on a copper substrate plate with an applied powder layer. The preheating temperature of the substrate plate and the beam size were varied to test different process conditions. The melt pool depths and widths were measured, and a relationship was elaborated. Based on these results, cube samples with discrete weld tracks on top were manufactured. The melt pool depth was measured and compared with the predicted melt pool depth to investigate the transferability of the elaborated relationship from the substrate to process conditions. It was found that with rising preheating temperature and for larger beam diameters at the same peak intensity, the weld width and weld depths increase. Furthermore, continuous weld tracks formed reliably in the keyhole welding regime. A good agreement between the weld depth of weld tracks on the substrate and the elaborated relationship was revealed. However, the weld tracks were shallower than predicted.

Numerical Simulation and Wear Resistance Property of Ni-Based Alloy Coating on the Surface of Ti-6Al-4V Substrate

Laser cladding is a new technology to improve the wear resistance or corrosion resistance properties of metal parts. A finite element model of laser cladding coating was established by numerical simulation technology. The temperature field distribution was studied and analyzed during the laser cladding process at three different scanning speeds and three different laser powers. A Ni-based coating was also fabricated on the substrate by a CO2 laser. The optimum parameters of the laser cladding were selected and compared with the melt pool depth and width of the Ni-based coating. Then, the cooling rate, temperature gradient, temperature and stress fields were calculated and analyzed. The growth mechanism of the crystal structure was analyzed by scanning electron microscope (SEM). The wear resistance of the Ni-based coating was measured by a friction and wear testing machine. The results showed that the optimal parameters were laser power 1600 W and scanning speed 3 mm/s. The temperature trends at different locations were similar. The calculated maximum residual stress was 0.157 GPa. The stress concentration appeared near the surface and both sides of the cladding layer. From the coating’s microstructure, it could be seen that it contained a large number of columnar dendrites, and the crystal size gradually decreased with the increase of cooling rate. The wear rates of the Ti-6Al-4V (TC4) substrate and the Ni-based coating were 6.98 mm3/(N·m) and 3.45 mm3/(N·m), respectively. The Ni-based layer had a low wear rate and good wear resistance, which is helpful to obtain good friction and wear resistance of TC4 substrates.

Multi-track, multi-layer cladding layers of YCF102: An analytical and predictive investigation of geometric characteristics

This paper proposes a prediction method of the geometric characteristic parameters of cladding layers with multi-track and multi-layer. YCF102 cladding layers with different number of tracks and layers were laser cladded on the AISI 1045 steel substrate. The changing trend of the geometric characteristic parameters was analyzed by a laser microscope, and the prediction method of clad height, clad width and melt pool depth with the different number of tracks and layers under various processing parameters were proposed respectively. The results show that the increase of the number of tracks and layers lead to the increase of clad height, clad width and melt pool depth. When cladding the sixth layer, the clad width and melt pool depth still show an increasing trend, but the increase rate of melt pool depth slows down. The average residual values between the calculated and measured values of clad height, clad width and melt pool depth are 280.01 μm, 158.90 μm and 8.00 μm respectively, which are only 5.69%, 4.01% and 1.72% of the corresponding measured values. It is indicated that the prediction method is reasonable. The proposed prediction method can be used to predict the geometric characteristic parameters of multi-track and multi-layer cladding layers. The application of the method can effectively save the processing time and cost of the laser cladding repair of machine parts.

Optimization of Butt-joint laser welding parameters for elimination of angular distortion using High-fidelity simulations and Machine learning

The high temperature gradient and high cooling rate induced in the butt-joint laser welding process generate a high residual stress, which can result in significant distortion of the final weld. Previous researches usually employed experimental approach to search for optimal parameters which is costly and time consuming. Additionally, there was several studies that used simulation to compute the distortion and to find optimal parameters, however, these studies neglect the effect of solid cooling rate and size of the heat-affected zone (HAZ) in the process of finding optimal parameters. Accordingly, the present study proposes a new systematic framework based on finite element (FE) simulations, regression analysis, and two artificial neural networks (ANNs) for determining the laser welding parameters (i.e., the laser power and welding speed) which minimize the angular distortion of the laser-welded butt joint and enhance the mechanical properties. In the proposed approach, to determine the variables of the selected double-conical heat source model for any combination of the laser power and welding speed is studied within the feasible design space. The variables are then imported into a three-dimensional COMSOL Multiphysics FE model to predict the melt pool dimensions, solid cooling rate, and size of the HAZ as a function of the laser power and welding speed. Finally, the processing map produced by the ANN is screened by four quality criteria based on the melt pool depth, angular distortion, cooling rate, and HAZ width, respectively, to determine the optimal laser welding parameters which minimize the angular distortion and maximize the mechanical properties. It shows that SS316L tensile test specimens fabricated using the optimal parameters have a tensile strength (662 MPa) which is higher than that of the base metal (660 MPa). Furthermore, the angular distortion of the butt-joint samples is less than 564 µm. Thus, the effectiveness of the proposed optimization procedure is confirmed.

Single-track investigation of additively manufactured mold steel with larger layer thickness processing: Track morphology, melt pool characteristics and defects

Large layer thickness is essential to increase building rate, but excessive layer thickness easily lead to defects which negatively affect component quality. The evolution of track morphology, melt pool characteristics and defects of additively manufactured mold steel by using various larger layer thickness (60 and 90 μm) while also varying the laser energy density was investigated systematically. Results indicated that the morphologies of single tracks shifted sequentially from discontinuousness to continuous irregularity and continuous regularity with the increasing line energy density at 60 μm layer thickness. The width and depth of melt pool increased substantially with the line energy density increased from 0.4 to 1.3 J/mm, while the height of melt pool was nearly constant. Correspondently, transition of the melt pool based on the depth-to-width ratio from balling to conduction and keyhole modes was observed. When the layer thickness increased from 60 to 90 μm, the average build percentage of the single tracks decreased by 43 % with increasing line energy density from 0.26 to 0.51 J/mm, and more spatters around the surface of tracks were found at 0.8 J/mm. Interestingly, the average width of the melt pool decreased by 37 μm, while the height of melt pool increased by 45 μm. More small spherical gas pores (5–15 μm) occurred in bottom of keyhole due to trapped vapor at high energy density. The decrease in melt width with increased layer thickness is most likely due to the weakened Marangoni convection. Finally, formation mechanism of lack-of-fusion defects and gas pores with larger layer thickness processing were analyzed based on melt pool characteristics, such as spheroidization, keyhole and spatters.

Preliminary Parametric Investigations into Macro-Laser Polishing of Laser-Directed Energy Deposition of SS 304L Bulk Structures

The higher surface roughness of laser-directed energy deposition (LDED)-built components necessitates advanced and sustainable surface quality enhancement techniques like laser polishing. In the present work, a parametric study involving experimental investigation and numerical analysis is conducted to determine the effect of macro-laser polishing on LDED-built SS 304L structures. A thermophysical model is developed to simulate the effect of laser power and scan speed on the melt pool depth of the LDED-built samples. The simulated melt pool depth is compared with experimental results and is found to be in good agreement. Further, the correlation between the melt pool depth and surface behaviour is studied based on shallow surface melting and shallow over-melting mechanisms. A maximum reduction in surface roughness from 21.3 µm to 9 µm (~57%) is achieved with laser polishing, and process parameters’ effect on the surface roughness is investigated. Scanning electron microscopy (SEM), energy-dispersive spectroscopy (EDS) mapping, and X-ray diffraction (XRD) are used to further characterize the laser-polished surface. SEM-EDS analysis shows that the segregation is more evident in laser-polished samples, while the XRD results indicate the absence of phase change during the process. This study paves the way to a greater understanding of the effect of macro-laser polishing on LDED-built SS 304L structures.

Multi-scale modelling for optimization of process parameters of laser powder bed fusion processed Inconel 718 surrogate part

This work uses an analytical and multi-scale thermo-mechanical numerical model to optimise the process parameter to prevent porosity while reducing the distortion in the surrogate part. Taguchi’s experimental design technique was used to identify the effect of process parameters on melt pool evolution. A dimensionless relative melt pool depth ratio (d*) ensures the extent of densification in the simulated part for various processing parameters. The Taguchi analysis outcome revealed an increase in d*, solid ratio, residual stress and distortion with increased laser power while a contrary effect with increased hatch spacing and scan speed. A strong interaction is observed for melt pool metrics and residual stress evolution. However, the interactions for residual stress and distortion evolution are more prominent, showing the influence of process parameters on energy density and thermal gradient evolution. The ANOVA analysis also shows a strong influence of process parameters on melt pool development and residual stress evolution. However, for solid ratio development, scan speed shows an insignificant effect. These observations from meso-scale analysis aids in process parameter selection for defect free part fabrication. A reduction in the VED reduced the magnitude of accumulated residual stresses in the surrogate component, thereby reducing the maximum distortion in the part. A decrease of 8.33 % in maximum distortion is observed for a ∼ 20 % reduction in the VED.

Effect of process parameters on microstructure and tribological properties of Ni60A/Cr3C2 laser cladding on 60Si2Mn steel

To improve the surface performance and service life of the arc-shaped nail teeth, a key component in pre-sowing film recovery machines, Ni60A/Cr3C2 composite coatings were deposited on the surface of 60Si2Mn steel using laser cladding technology. A numerical simulation of the temperature field was conducted for different process parameters based on the ANSYS secondary development language APDL and life–death element technique. X-ray diffraction (XRD), scanning electron microscopy (SEM), three-dimensional morphometry, and X-ray photoelectron spectroscopy (XPS) were also used to analyze the microhardness, microstructure, elemental distribution, and wear resistance of the composite coatings, respectively, to investigate the optimum combination of laser coating parameters. Results show that the temperature field demonstrates a “comet-like” distribution, forming an elliptical melt pool. The average error between melt pool depth and experimental results under this model is 5.1 %. The process parameters affecting the quality of the coating are presented in order of priority: scanning rate, powder feed rate, and laser power. Composite coatings exhibit precipitation of new phases such as γ-Ni-based, NiO, M7C3 (M = Cr, Mn), and M23C6 type. The optimum combination of laser cladding parameters (T6) is 1800 W laser power, 5 mm/s scanning rate, and 8 g/min powder feed rate. T6 coating exhibits a good metallurgical bond with the substrate and a microhardness of 902 HV0.1. Wear of the T6 coating is mainly in the form of abrasive wear, while wear of the base material is in the form of severe adhesive and abrasive wear. The wear and surface roughness of the T6 coating is only 13 % and 39.5 % than those of the 60Si2Mn base material, and the wear depth is reduced by 77 %. This study can provide a reference value for the use of surface technology to enhance the comprehensive performance of agricultural machinery and equipment.

A framework for the robust optimization under uncertainty in additive manufacturing

This paper introduces a conceptual framework for the robust optimization under uncertainty in the directed energy deposition (DED) of M4 High-Speed Steel. The goal is to identify optimal process parameters for robust manufacturing of printed parts with a stationary melt pool depth and low consumed energy under uncertainty within the multiple layers of a bulk sample. To increase the computational efficiency, a deep learning-based surrogate model is built using the training data generated by a validated high-fidelity DED two-dimensional FE model. The robustness of the optimized result is verified using the Monte-Carlo method and compared with experiments and two other deterministic approaches. Furthermore, we conduct a global sensitivity analysis, which indicates that among six uncertain input variables, the thermal conductivity and the convection have the most significant impact on the melt pool depth variation. This study shows the promising possibilities of the presented framework in optimizing the DED process.