Melt Pool Temperature Research Articles

In Situ Monitoring of Optical Emission Spectra for Microscopic Pores in Metal Additive Manufacturing

AbstractQuality assurance techniques are increasingly demanded in additive manufacturing. Going beyond most of the existing research that focuses on the melt pool temperature monitoring, we develop a new method that monitors the in situ optical emission spectra signals. Optical emission spectra signals have been showing a potential capability of detecting microscopic pores. The concept is to extract features from the optical emission spectra via deep auto-encoders and then cluster the features into two quality groups to consider both unlabeled and labeled samples in a semi-supervised manner. The method is integrated with multitask learning to make it adaptable for the samples collected from multiple processes. Both a simulation example and a case study are performed to demonstrate the effectiveness of the proposed method.

Numerical and analytical investigation on meltpool temperature of laser-based powder bed fusion of IN718

Prediction of meltpool features in Laser-Based Powder Bed Fusion (LB-PBF) is a complex non-linear multiple phase dynamic problem. In this investigation, numerical simulations and analytical models are offered to predict meltpool temperature and to provide a methodology to estimate melt track quality. By determining the meltpool temperature, different rheological phenomena including recoil pressure can be controlled. Recoil pressure is known to drive the keyhole and conduction modes in LB-PBF which is an important factor to qualify the melt track. A numerical simulation was carried out using Discrete Element Method (DEM) with a range of process parameters and absorptivity ratios; allowing observation of the variation of meltpool temperature and free surface morphology, as calculated by the volume-of-fluid (VOF) method. A spatially thermophysical-based analytical model is developed to estimate meltpool temperature, based on LB-PBF process parameters and thermophysical properties of the material. These results are compared with experimentally observed meltpool depth for IN718 specimens and found to have a good accuracy. The numerical and analytic results show good agreement in the conduction mode to estimate the meltpool temperature and related phenomena such as recoil pressure to control the melt track and layering quality. The analytical model does not accurately predict the keyhole mode which may be explained by evaporation of chemical elements in the examined material.

In situ alloying based laser powder bed fusion processing of β Ti–Mo alloy to fabricate functionally graded composites

β-titanium (Ti) alloys combined with high strength and good ductility have attracted extensive research interest for use in many advanced industrial applications. In this study, laser powder bed fusion (LPBF) parameters were firstly optimised to fabricate highly dense metastable β Ti–12wt% Mo alloy with largely homogeneous structure from low-cost elemental powders. When the laser area energy density (AED) increased to 4 J/mm2 using the simple scan mode, the refractory Mo powder melted and dissolved into the Ti matrix due to the increased melting pool temperature and increased cycles of remelting. However, when the same high AED was used in the chess scanning mode, keyhole-induced defects emerged along the island boundaries. The laser beam delay (LBD) and island spacing (IS) settings were then optimised to eliminate keyhole defects. Additionally, it is found that using low and high AED (e.g. 1.6 and 4 J/mm2 respectively), the builds show significantly different microstructure and mechanical properties. In view of this, a Ti–Mo functional gradient composite (FGC) was fabricated via alternating AEDs of 1.6 and 4 J/mm2 layer-wise. A gradient distribution of α” phase with varied size and quantities across the designed layer boundaries was produced. The Ti–Mo FGC possessed a high compressive yield strength of 1173 (±15) MPa and improved strain hardening capacity. The developed approach demonstrated the potential for the fabrication of FGCs using an in situ alloying based LPBF.

Energy reconstruction and metallurgical characteristics in 316L NG-LWFW with assisted wire wobbling by numerical and experimental analysis

A novel, new approach of narrow gap laser wire feeding welding (NG-LWFW) with electromagnetic assisted wire wobbling is proposed to solve incomplete sidewall fusion and improve solidification morphology. The energy distribution and fusion zone formation during multi-pass NG-LWFW process are clarified. The melt pool is driven into sidewall position and the melted metal accumulates towards that side groove, leading to the increase of sidewall penetration with concave bead profile. The simulation results show that the wire wobbling obviously reduces peak temperature of melt pool and increases sidewall temperature. The heat affected zone (HAZ) width reduced, and a finer and diversity weld microstructure is presented. As multi-pass welding proceeds, the former weld is heat-treated by the latter weld, leading to the re-melting of columnar grains and increasing diversity of interlayer grains. The columnar crystal zone is refining and a narrower dendrite spacing of δ-ferrite phase is found with deposited filling layers. The micro-hardness of weld centerline shows an overall increase trend and the impact energy increases with denser dimples from the bottom layer to upper layer.

Closed-loop deposition of martensitic stainless steel during laser additive manufacturing to control microstructure and mechanical properties

A closed-loop laser additive manufacturing (CLAM) system was developed to monitor and control layer-by-layer fabrication via directed energy deposition (DED). The well-modified laser additive manufacturing (LAM) package utilizes a thermal infrared camera and thermal image processing algorithms to extract and control the thermal dynamics of the process, including (i) cooling rate, (ii) melt pool temperature, and (iii) heating rate. Martensitic stainless steel of grade S410-L was used to cumulatively deposit 5 layers on top of each other at different processing conditions without closed-loop control. Significant deviations of cooling rate and melt pool temperature are observed within different layers of a thin-wall structure without feedback control in the open-loop state. To assess the effects of cooling rate control on final LAM properties, three samples with different setpoint cooling rates of 550, 1100, and 1750 °C/s are tested under closed-loop conditions. Microstructural and microhardness maps and profiles of open-loop and closed-loop samples reveal highly uniform and controlled hardness profiles when closed-loop control is applied, in contrast to highly varying properties in samples produce in open-loop operation. The current research results illustrate how one can achieve controlled microstructural characteristics and mechanical properties through real-time control of the cooling rate during the LAM process.

Effect of Heat Treatment and Electric Discharge Alloying on the Tribological Performance of Selective Laser Melted AlSi10Mg

Abstract Selective laser melting (SLM) is an emerging additive manufacturing (AM) technology for fabrication of complex lightweight components along with improved mechanical properties. However, the properties are highly influenced by the continual heating and cooling during deposition, variation in local temperature, size, and shape of melt pool, and solidification growth rate. Therefore, postprocessing is very often required to control various properties of additive manufactured components. The present work investigates the influence of various postprocessing methods such as heat treatment and electric discharge alloying (EDA) on ambient and elevated temperature wear behavior of selective laser melted AlSi10Mg alloy and compared with its tribological behavior with cast AlSi10Mg. The dry wear tests were conducted using a pin on disk (POD) tribometer with EN-31 as counter body. The EDA treated SLM AlSi10Mg showed the least wear-rate and coefficient of friction (COF) at both ambient and elevated temperatures (1.05 × 10−4 mm3/Nm and 0.434 and 3.12 × 10−5 mm3/Nm and 0.531, respectively) due to its higher hardness (189.8 HV) as compared with other samples. The wear-rate and COF of cast specimen are found to be highest among all specimens at both ambient and elevated temperatures (1.34 × 10−4 mm3/Nm and 0.528 and 4.49 × 10−5 mm3/Nm and 0.724, respectively). Lower wear-rate and higher COF are observed at elevated temperature due to the excessive formation of wear-resistant oxides (Al2O3, SiO2, and MgO) and glaze layers for all samples compared with ambient temperature wear behaviors of its counterparts. Abrasive wear, adhesive wear, oxidation wear, and surface delamination are the prominent wear mechanisms observed for ambient and elevated temperatures for all the specimens.

On the role of process parameters on meltpool temperature and tensile properties of stainless steel 316L produced by powder bed fusion

This research aims to identify how meltpool temperature is determined by process parameters in Laser-Based Powder Bed Fusion (LB-PBF) and the effect of meltpool temperature and heat treatment temperature on microstructure and tensile properties. The study illustrates how crystallographic features in 316L stainless steel were developed in response to the meltpool temperature and induced energy density of LB-PBF manufacture, and by post manufacture heat treatment. For this research, 25 samples based on a Taguchi Design of Experiments (DoE) with five parameters over five levels were printed. To improve precision, the DoE was repeated three times and a total of 75 samples were produced. A thermophysical-based analytical model was developed to measure the meltpool temperature and subsequently surface tension of the liquid in melting zones. Then, a statistical method was used to identify the effective process parameters in tensile properties including ultimate strength, fracture strain and toughness. The microstructural evaluation and crystallographic features were presented to identify the governing mechanisms for the discussed phenomena. This result verifies that the meltpool temperature is a driving factor for the microstructural evolution and observed crystallographic features. Results showed that samples with lower meltpool temperatures have smaller grain sizes, superior strength and toughness properties. The crystallographic analysis showed the weak texture and anisotropic properties are dominant by the preferred orientation growth. The geometrically necessary boundary values were also found to be a driving factor for fracture strain. The originality of this paper is identifying the effect of process parameters on meltpool temperature using an analytical model that is developed based on the thermophysical properties of the feedstock. Characterizing the effect of meltpool temperature in crystallographic features are also another contribution of this paper.

Pulsed laser welding and microstructure characterization of dissimilar brass alloy and stainless steel 308 joints

The laser welding of stainless steel and brass is difficult due to their different physical properties, including thermal conductivity and melting point. In this study, the laser welding of stainless steel 308 and brass was experimentally investigated. To systematically investigate the effects of different thermophysical properties, including the heat transfer rate, thermal conductivity, and laser beam interaction in relation to the beam absorption coefficient, the temperature of the melt pool zone was measured while changing welding conditions. Different temperature gradients were obtained by changing different parameters including the laser peak power, nozzle distance, and laser beam deviation. The results showed that an asymmetric melt pool formed, to which the melting of brass made a higher contribution. Due to the lower melting temperature and the higher thermal conductivity of brass, the measured temperature and the melt size of brass were higher. The microstructure of the melt pool consisted of intermetallic compounds. Variation of melt pool size (width and depth) and temperature could be affected by laser peak power and reductions in the nozzle distance than the other parameters. For instance, 1 mm increase in the nozzle distance reduced the melt pool zone temperature by approximately 20 °C for stainless steel. Furthermore, an increase in the laser peak power raised the maximum measured temperature to about 225 and 160 °C for brass and stainless steel sides, respectively. At the same time, the amount of temperature reduction at a period of 20 s for stainless steel is almost half that of brass.

FEM analysis of thermal and residual stress profile in selective laser melting of 316L stainless steel

High temperature gradients and resulting large residual stresses are among the most challenging areas of selective laser melting. In order to control these issues, proper understanding of the influence of process parameters on thermal as well as residual stress profile during selective laser melting is of paramount importance. Despite the large number of parametric investigations regarding stress profiles, a lot is yet to be understood especially for 316L stainless steel. Therefore, a 3D FEM model is established in this work to investigate the parametric influence on thermal and residual stress profile in a realistic multi-track multi-layer SLM process. Influences of process parameters on thermal and stress profile were investigated through the experimentally validated FEM model. It was observed that an increase in laser power or a decrease in the scanning speed results in an increased maximum melt pool temperature. Whereas, increasing laser power or scanning speed both increases cooling rates and residual stresses. An increased maximum melt pool temperature and decreased residual stress was observed by increasing the hatch spacing. Residual stress profile was observed to be symmetrical along the scanning direction and non-uniform along the direction perpendicular to the scanning direction. A relatively small residual stress at the top surface and maximum residual stress inside the component around the interface region of powder bed and baseplate was observed.

Microstructure and properties of parts manufactured by directed energy deposition of water-atomized low-alloy steel powders

Abstract The feasibility of using water-atomized low-alloy steel powders [Fe-0.11C-0.59Mn-0.43Si-0.45Cu-0.3Cr-0.45Ni-0.15Mo-0.05 V (wt.%)] for additive manufacturing using a laser-based directed energy deposition (DED) process was evaluated. Different processing parameter combinations were assessed to arrive at optimum parameters to fabricate different sample geometries, which were characterized in terms of microstructures and mechanical properties. Selected samples were also evaluated after homogenization at 1075 °C for 10 min. The results show that DED of water-atomized low-alloy steel powders is possible, yielding unique non-directional and refined microstructures with fine spherical oxide particles dispersed throughout the matrix. The maximum tensile strength and total elongation achieved in the as-built condition were 315 MPa and 3.4%, respectively, which improved to 350 MPa and 10.3% after heat treatment. The presence of a sizeable fraction of oxide nanoparticles, which presumably originated from the oxide layer of the water-atomized powder, enabled retention of microhardness/strength after heat treatment despite grain coarsening. However, the porosity and loss of C and Mn were found to be the major concerns with water-atomized powder. We believe that inconsistent powder flow due to the irregular morphology of the powders is the primary cause of the porosity. On the other hand, the high laser absorptivity of the powders due to their rough surface texture and surface oxides could be responsible for the loss of alloying elements due to excessive powder and/or melt pool temperatures. Although the results of our study are quite encouraging, powder characteristics like size and/or morphology need to be further optimized to exploit the full potential of this additive manufacturing methodology.

Implementation of Abaqus user subroutines and plugin for thermal analysis of powder-bed electron-beam-melting additive manufacturing process

Electron beam melting (EBM) is a metal powder bed fusion additive manufacturing (AM) technology that is widely used for making three-dimensional (3D) objects by adding materials layer by layer. EBM is a very complex thermal process which involves several physical phenomena such as moving heat source, material state change, and material deposition. Conventionally, these phenomena are implemented using in-house codes or embedding some user subroutines in commonly used commercial software packages, like Abaqus, which generally requires considerable expertise. Fortunately, recent versions of Abaqus offer a new plugin tool, AM Modeler, which provides a rather new and user-friendly method for performing additive manufacturing process simulation. In this work, taking Ti-6Al-4V as the particular example, we present all the details of the finite element (FE) implementation of both Abaqus user subroutines and AM Modeler plugin for thermal analysis of EBM additive manufacturing process. The melting pool shape and temperature profiles were predicted and verified against existing literature data. A 3D FE model was also developed to capture the heat transfer features in a real manufacturing process for printing a particular 3D object, “AM” characters, validating the capability of the proposed methods. To facilitate future design and thermal analysis of EBM process and to promote the use of AM Modeler plugin, the source codes including Abaqus user subroutines DFLUX and UMATHT as well as Python scripts for implementing AM Modeler are made available and could be downloaded from https://github.com/Dr-Ning-An/ebm-abaqus.

The Thermo-Mechanical Coupling Effect in Selective Laser Melting of Aluminum Alloy Powder.

In the selective laser melting process, metal powder melted by the laser heat source generates large instantaneous energy, resulting in transient high temperature and complex stress distribution. Different temperature gradients and anisotropy finally determine the microstructure after melting and affect the build quality and mechanical properties as a result. It is important to monitor and investigate the temperature and stress distribution evolution. Due to the difficulties in online monitoring, finite element methods (FEM) are used to simulate and predict the building process in real time. In this paper, a thermo-mechanical coupled FEM model is developed to predict the thermal behaviors of the melt pool by using Gaussian moving heat source. The model could simulate the shapes of the melt pool, distributions of temperature and stress under different process parameters through FEM. The influences of scanning speed, laser power, and spot diameter on the distribution of the melt pool temperature and stress are investigated in the SLM process of Al6063, which is widely applied in aerospace, transportation, construction and other fields due to its good corrosion resistance, sufficient strength and excellent process performance. Based on transient analysis, the relationships are identified among these process parameters and the melt pool morphology, distribution of temperature and thermal stress. It is shown that the maximum temperature at the center point of the scanning tracks will gradually increase with the increment of laser power under the effect of thermal accumulation and heat conduction, as the preceded scanning will preheat the subsequent scanning tracks. It is recommended that the parameters with optimized laser power (P = 175–200 W), scanning speed (v = 200–300 mm/s) and spot diameter (D = 0.1–0.15 mm) of aluminum alloy powder can produce a high building quality of the SLM parts under the pre-set conditions.

Advances in the understanding of the annular laser beam wire cladding process

This paper experimentally studies the influence of conventional process parameters (laser beam power, P, workpiece, and wire feeding velocities), and the nominal workpiece irradiation proportion WIPN parameter, on ALB-WC process stability and generated clad geometry. The study was based on an extensive set of single clad experiments using stainless steel AISI 316L wire of 0.6 mm diameter and a stainless steel AISI 304 workpiece. The results show that the laser beam power required for a stable process increases non-linearly with WIPN, and this relationship can be described by quadratic regression. An increase in either the workpiece or wire feeding velocity respectively causes a reduction or enlargement of the stability region in the P-WIPN stability diagrams. The influences on stability diagrams are explained by means of the increase in the nominal WIPN to the effective WIPE, as a result of the clad height generated. With respect to clad geometry, it was shown that an increase in the WIPN causes a decrease in the dilution ratio, which is linearly correlated with the workpiece and melt pool temperatures, while clad width and height are not influenced by the WIPN. The paper further explains the complex influences of the process parameters observed on process stability and generated clad geometry using a phenomenological description of the ALB-WC process at a low and high nominal WIPN. For this purpose, heat transfer, Marangoni flow, and laser beam reflections are considered qualitatively and supported by IR and visual images of the wire-end and melt pool as well as the workpiece and melt pool temperatures.

A new method for improving the forming quality in laser additive manufacturing process

The melt pool temperature plays a very important role in the laser additive manufacturing (LAM) process which has been widely used in different industries. In this process, simple laser scan strategies and constant laser power levels are usually used, which will elicit variations in the melt pool temperature and ultimately affect the forming quality such as the cracks and pores. In this paper, a new method for improving the forming quality was presented, which calculates the local thermal distribution dynamically by analyzing the laser scanning path included in a G-code file. A factor called the thermal distribution factor (TDF) was used to adjust the laser power. By inserting the adjusted laser power into a modified G-code file, the melt pool temperature can be controlled indirectly. This method can help to control the melt pool temperature and improve the forming quality at the lowest cost. An experimental set-up was designed to verify this method and the results were analyzed. Results indicated that the cracks and pores of the part were significantly reduced by using the modified G-code file. Besides, the drawbacks and the future potential improvements of this method were discussed.

Digitalization of directed energy deposition process through a multidirectional height monitoring sensor system

Industrialization of directed energy deposition (DED) process relies on the development and implementation of a quality assurance system that makes the process reliable and repetitive. One of the geometrical characteristics that needs constant monitoring and control during any DED process is the height of the deposited layer. In this paper, a multidirectional layer height monitoring system relying on the laser triangulation principle that has been developed for real-time direct measurement, in contrast to the majority of the sensor systems that derive layer height based on melt pool temperature, is presented. Existing commercial laser triangulation systems are unidirectional, which hinder the flexibility of the material handling system, like industrial robot, in most of the cases. An additional challenge lies on the integration with the deposition head due to its form, size, and extreme working conditions. Therefore, a novel configuration of laser triangulation sensors that enable multidirectionality and limit shadowing effect has been developed at Fraunhofer IAPT. In this paper, the development of such a sensor with a focus on the mechanical design has been presented. The sensor is designed for operation at high temperature and harsh environmental conditions close to the build zone. The developed sensor has been successfully tested for the powder-based laser metal deposition system. In-process layer height measurement and control are possible with this sensor, thereby enabling digitalization and quality assurance during the build process. 3D point cloud generated using this sensor can be used for the dimensional deviation measurement and also for downstream processes like machining. Such an in-process quality measurement and assurance system negates a postmeasurement step, thereby saving cost and time and ensuring quality.

Thermo-physical modeling and experimental validation of core-shell nanoparticle fabrication of nickel-titanium (nitinol) alloy

Nickel-titanium (nitinol) is a biocompatible shape memory alloy with hyperelasticity and biocompatibility property. A three-dimensional heat diffusion model was used in conjunction with phase explosion theory to explain the formation of core-shell nanoparticles of nitinol. The diffusion model was used to evaluate the melt-pool temperature in the ablation region, while the phase explosion theory was used to estimate the thermophysical expansion and subsequent fragmentation of molten and vaporized nitinol. The mathematical parameters were formulated in conjunction with the experimental results of pulsed laser ablation of core-shell nitinol nanoparticles of various sizes. The model was further employed to predict the sizes of the nanoparticles produced at other higher fluences and the results were experimentally validated.

On-line melt pool temperature control in L-PBF additive manufacturing

Laser powder bed fusion is a promising additive manufacturing technology which has enabled the fabrication of complex-shape, custom-designed, and cost-effective parts with no need for expensive tools and dies. Despite the numerous advantages of this technology, inconsistency in the microstructure and, consequently, the mechanical properties of the fabricated components in the building direction makes it quite challenging to obtain uniform parts in the as-built state. This issue originates from the layer-by-layer melt pool temperature variation caused by the layer-wise nature of this process when a fixed set of process parameters is applied. Accordingly, a layer-wise melt pool temperature control system is beneficial in manipulating the process parameters and therefore adjusting the melt pool temperature. In this study, three different controllers, namely, simple proportional (P), adaptive P, and sliding mode, were designed to control the melt pool temperature in the building direction for Inconel 625 superalloy. An analytical-experimental model was introduced to evaluate the performance of controllers through simulation. A monitoring system having a two-color pyrometer was used to online monitor the temperature for use by the controllers as a feedback signal. The microstructure and microhardness of the final products were evaluated prior to and after employing the melt pool temperature controllers. Compared to the scenario with constant process parameters, the implementation of these controllers led to improved microhardness and microstructure uniformity, resulting from the reduced variation in the primary dendrite arm spacing. The lessons learned from this study can assist in the fabrication of functionally graded materials with engineered microstructures.

A Hybrid Deep Learning Model for Layer-Wise Melt Pool Temperature Forecasting in Wire-Arc Additive Manufacturing Process

Melt pool temperature contains abundant information on metallurgical and mechanical aspects of products produced by additive manufacturing. Forecasting melt pool temperature profile during a process can help in reducing microstructural porosity and residual stresses. Although analytical and numerical models were reported, the performance of these are questionable when applied in real-time. Hence, we developed data-driven models to address this challenge, for continuous forecasting layer-wise melt pool temperature using a hybrid deep learning technique. The melt pool temperature forecasting by the proposed CNN-LSTM model is found to be better than other benchmark models in terms of accuracy and efficiency. The model results have shown that combining CNN and LSTM networks can extract the spatial and temporal information from the melt pool temperature data. Further, the proposed model results are compared with existing statistical and machine learning models. The performance measures of the proposed CNN-LSTM model indicate a greater potential for in-situ monitoring of additive manufacturing process.

Microstructure and Wear Behavior of NbC-Reinforced Ni-Based Alloy Composite Coatings by Laser Cladding

This work aims to evaluate the influence of laser power and reinforcement feeding rate on the microstructure, hardness, and wear behavior of NbC-reinforced Hastelloy C276TM alloy composite coatings. From a dual feeding system, one-step metal-matrix composite coatings were deposited with 10, 30, and 50% mass feeding of NbC powder with laser powers of 1.5 and 3.0 kW. Coatings deposited with 1.5 kW and 30% NbC showed some degree of porosity due to a combination of NbC feeding rate and melting pool temperature. Laser power and NbC feeding rate altered the melting efficiency and substrate burn-in shape, remarkably influencing the dilution. The composite microstructure was comprised of Ni-γ (FCC) dendrites with interdendritic network carbide which, in turn, ranged from lamellar-like M6C to blocky-like conjugated MC-M23C6 carbide. Primary petaloid-like MC [Nb] carbide was formed from a 30% reinforcement rate on, whilst a considerable number of unmelted particles was observed for 50%. The higher the reinforcement feeding rate, the higher the carbide fraction and better wear performance of low heat-input specimens. Synthesis with high heat-input sheds a light on the deleterious effect of the dilution and is not an option to enhance the wear performance.

DLAM: Deep Learning Based Real-Time Porosity Prediction for Additive Manufacturing Using Thermal Images of the Melt Pool

This paper presents an investigation of the rapid variations in the temperature of metal melt pool for Additive Manufacturing (AM) processes. The melt pool is created by scanning a high-power laser beam across a metal powder bed. Rapid heating and cooling processes are involved in the layer-by-layer fabrication of the metal part. Recent advances in Machine Learning and Deep Learning algorithms provide efficient ways to analyze large sets of data in search of correlations that would otherwise be extremely time-consuming. The use of Machine Learning and Deep Learning algorithms to understand temperature variations in AM fabrication process will allow to predict the formation of porosity before it occurs. The objective of this research is to advance the AM technology using enhanced Deep Learning techniques to provide in-situ analysis of the melt pool temperature that can lead to a reliable manufacturing of Three-Dimensional (3D) metal parts/components. In specific, Deep Learning based porosity prediction for Additive Manufacturing (DLAM) methods have been proposed. In DLAMs, several state-of-the-art Deep Learning algorithms such as Convolutional Neural Networks (CNN) using transfer learning, and Residual-Recurrent Convolutional Neural Networks (Res-RCNN) are proposed for effectively performing the end-to-end porosity prediction in real-time using thermal images of melt pool. Experimental results, in this research, show that the Res-RCNN has an overall accuracy of 99.49% and inference time of $8.67ms$ , and the Res-RCNN outperforms other baseline models. The Res-RCNN’s recursive architecture allows the network to view each input image multiples times and at varying feature levels, which enables a slight boost in porosity prediction accuracy over the commonly used transfer learning CNN models.