Toward resilient energy assets: comparative evaluation of additive manufacturing technologies for critical spare parts

The energy sector, particularly oil and gas operations, relies heavily on long-service, high-reliability components that are often custom-fabricated or sourced from legacy supply chains (Urciuoli et al., 2014). Many of these components face issues of obsolescence, long lead times and unavailability of original tooling or suppliers, especially in remote or offshore installations (Wahab et al., 2018; Ferreira et al., 2023). These challenges are compounded by the increasing need for agile maintenance solutions and enhanced supply chain resilience to avoid downtime in critical systems such as chemical pumps, valves and rotating equipment (Tartoor et al., 2020; Ekram, Elmesmary and Sakr, 2024). Additive manufacturing (AM) offers a transformative alternative for reproducing and improving such components (Sireesha et al., 2018; Rahito, Wahab and Azman, 2019). With capabilities including design freedom, on-demand production and reduced material waste, AM is increasingly viewed as a potential solution to extend the operational life of aging assets through part remanufacturing and functional replacement (Kostidi and Nikitakos, 2019; Kanishka and Acherjee, 2023; Alabtah, Alkhouzaam et al., 2025).

In particular, metal AM techniques such as Laser Powder Bed Fusion (LPBF) and Bound Metal Deposition (BMD) have garnered attention for their ability to fabricate geometrically complex, mechanically robust parts (King et al., 2015; Mahmood et al., 2023; Badoniya et al., 2024; Fatemeh Nabavi, Farshidianfar and Dalir, 2025). LPBF is known for its high resolution and strong mechanical performance, making it suitable for parts requiring close tolerances and wear resistance (Chowdhury et al., 2022; Narasimharaju et al., 2022). BMD, on the other hand, is gaining popularity as a more accessible, safer and potentially lower-cost process for fabricating medium-complexity structural parts (Iacopo et al., 2022; Mechter, Mace and Kerbrat, 2022). Over the past decade, considerable research has focused on characterizing the microstructural and mechanical behavior of parts produced using these technologies (Di Pompeo et al., 2023; Ansari, Ghassan Alabtah and Khraisheh, 2024; Nabavi, Dalir and Farshidianfar, 2024). LPBF studies have demonstrated the sensitivity of part quality to process parameters such as laser power, scan speed and hatch spacing, with a growing emphasis on defect control through melt pool simulation and thermal modeling (Sarkar, Kapil and Sharma, 2024; Tan and Spear, 2024). BMD, a relatively newer approach, has seen increased interest for structural applications, particularly in stainless and tool steels (Alabtah, Chihi et al., 2025), where research has focused on shrinkage behavior, sintering conditions and achievable mechanical properties (Gabilondo et al., 2022; Bellezze et al., 2023). At the same time, reverse engineering and design tools; particularly high-resolution 3D scanning, generative design (GD) and finite element analysis (FEA); have matured to enable part reconstruction and functional optimization. These technologies allow legacy components to be not only replicated, but potentially improved in terms of weight, stress distribution, or manufacturing efficiency (López and Vila, 2021; Karlsen and Ratnayake, 2023).

Despite this progress, the vast majority of AM research remains constrained to coupon-level characterization, parametric optimization, or simulation of simplified geometries (Gardner, 2023; Mohr, Altenburg and Hilgenberg, 2023). Studies often evaluate AM technologies in isolation, without direct comparison under shared design and testing conditions (Khan and Riccio, 2024; Shen et al., 2025). More critically, relatively few investigations extend their scope to part-level validation, where an additively manufactured component is evaluated not only through laboratory testing, but also for dimensional fidelity, functional performance and real-world service integration (Dordlofva, 2020; Obilanade, Törlind and Öhrwall Rönnbäck, 2025). This limits the ability to translate material- and process-level findings into qualification decisions for fit- and function-critical components with complex geometries and tight tolerance requirements. As a result, industrial stakeholders still lack guidance on selecting suitable AM processes for specific application contexts, particularly when part complexity, mechanical requirements and processing constraints intersect (Cardeal, Ribeiro and Leite, 2025). Moreover, the integration of digital reconstruction, simulation-informed process planning, mechanical evaluation and functional verification into a unified, application-driven workflow remains rare in the AM literature (Novak, Ren and Vesenjak, 2025; Zhang et al., 2025).

This study aims to address these gaps through a comprehensive, experimentally validated case study involving the reproduction of a reverse-engineered eccentric gear used in a chemical demulsifier pump (P-5636A/B) for oil and gas operations. The original component, manufactured by casting, was digitally scanned and reconstructed in CAD. GD was performed using Autodesk Fusion 360 to explore lightweight structural configurations that reduce material usage while preserving the load-bearing capacity of the gear, and to ensure the geometry could be feasibly manufactured using BMD. Following GD, a static structural FEA was performed in ANSYS Workbench to verify that the GD part preserved mechanical integrity under the anticipated pump duty. Two process and material pairings were selected to reflect realistic deployment options and to satisfy end-user functional requirements: LPBF with a cobalt-chrome (CoCr) alloy, and BMD with 4140 chromoly steel.

To de-risk build trials for the intricate geometry, ANSYS Additive Print 2024 R1 was used to predict distortion and displacement for various build orientations of LPBF-CoCr parts; the lower-risk orientation was then manufactured. Process-matched coupons were produced for tensile testing (ASTM E8/E8M) and microhardness (ASTM E92); comparable coupons were extracted by Electrical Discharge Machining (EDM) from the original gray cast iron gear to benchmark baseline performance. The printed gears were subjected to dimensional conformity assessment using a high-resolution 3D scanner. A scan-to-CAD deviation analysis was performed to quantify deviations from the as-designed geometry, with particular attention to fit-critical features. Prior to deployment, gear mesh quality was verified by a gear contact test conducted consistent with API 677 procedures, using controlled alignment and applied load to evaluate contact localization and load distribution across the tooth face. The resulting patterns indicated acceptable mesh alignment without edge loading. The additively manufactured GD eccentric gear was installed in the operating pump and has demonstrated reliable operation in field service to date. The novelty of this work lies in its integration of design, simulation, process comparison and field validation into a single, application-focused study. By moving beyond coupon-level reporting, the study provides a part-level, performance-oriented workflow to support process selection and qualification for critical spare-part replacement in energy assets. This study addresses the following research questions:

To enable the functional reproduction of a legacy energy-sector component using AM, a structured methodology was established that integrates digital reconstruction, design optimization, process planning, fabrication, and validation. Figure 1 summarizes the conceptual framework of the study, structuring the approach into sequential stages from reverse engineering through validation and field deployment. The workflow begins with reverse engineering of the cast gear through high-resolution 3D scanning and CAD reconstruction, followed by GD and structural FEA to ensure that optimized geometries maintain mechanical integrity under service loads. Prebuild process simulation is then employed to predict distortion and identify suitable build orientations prior to fabrication.

Two AM methods, LPBF with CoCr and BMD with 4140 steel, were selected to reflect realistic industrial deployment options, with subsequent post-processing applied to achieve final part quality. The selection was informed by functional requirements typical of legacy spare parts in energy-sector equipment, for which tolerance-critical interfaces and wear-related performance are key determinants of service readiness. LPBF-CoCr was selected to represent a high-precision route with well-established industrial maturity for producing dense parts with fine features and tight dimensional control; in addition, cobalt-chromium alloys are commonly used in demanding service environments where high hardness and wear resistance are desirable, and they provide strong corrosion resistance in chemically aggressive conditions. In contrast, BMD-4140 was selected to represent an accessible steel-based route aligned with maintenance-driven deployment: 4140 is widely used in rotating machinery and power/energy applications due to its strength-toughness balance, heat-treatability, machinability and broad availability, and it is compatible with bound-metal feedstock formats used for lower-barrier implementation. These pairings therefore represent two feasible process-material routes for legacy spare parts; the comparison is intended to inform route-selection trade-offs and qualification evidence requirements rather than to provide a like-for-like alloy substitution for the original cast material. Manufactured gears were evaluated through standardized mechanical testing, dimensional verification and functional inspection via API 677 gear contact-pattern testing before field installation under real operational conditions.

The legacy component investigated in this study is an eccentric wheel gear from a chemical pump used in oil and gas operations. Originally fabricated by casting, the part lacked any accompanying CAD documentation or digital design files. To reproduce the component for AM, a reverse engineering workflow was implemented to digitally capture its geometry and fit-critical features. High-resolution 3D scanning was conducted using the Artec Space Spider, which offers point accuracy up to 50 μm, suitable for capturing intricate details such as gear teeth, shaft interfaces and mounting surfaces. Multiple passes were acquired from various orientations to ensure complete surface coverage. The resulting point cloud was processed in Artec Studio to produce a watertight, topologically consistent mesh. This mesh was subsequently used as the basis for reconstructing a CAD model that accurately preserved the functional geometry of the original component. Figure 2 illustrates the reverse engineering workflow, from the original cast part, followed by scanning, mesh generation and final CAD reconstruction. The resulting digital model served as the foundation for downstream GD, structural verification and process planning prior to AM.

The original gear utilized for reverse engineering was a service component extracted from operation. Prior to 3D scanning, the gear was visually inspected and measured using a calibrated digital caliper to assess its wear condition. No significant deformation, tooth fracture, or abnormal pitting was observed, although minor polishing consistent with operational use was present. To mitigate the influence of localized wear during CAD reconstruction, Geomagic Design X^TM was employed to extract best-fit geometric primitives rather than directly copying raw surface mesh irregularities. Cylindrical regions such as hub diameters and the bore were generated using best-fit cylinder tools, while tooth geometry was reconstructed from multiple cross-sectional slices to ensure dimensional consistency across several teeth. Key dimensions, including circular pitch and keyway geometry, were in addition verified using manual measurements. This combined approach ensured that the reconstructed CAD model represented the intended nominal geometry rather than localized service wear.

AM enables internal architectures that direct material along primary load paths, allowing substantial weight reduction without sacrificing functional integrity. In this study, this capability is used to produce a service-ready eccentric gear. For BMD, through it is dominated by the debinding-sintering cycle, complete debinding of the fully solid gear [Figure 3(a)] was deemed impractical as the Desktop Metal Fabricate™ software predicted an unrealistically long primary debinding cycle of 655 h and a total thermal cycle of 708 h 56 min, straining furnace capacity and risking property degradation from prolonged thermal exposure. A 30% infill alternative [Figure 3(b)] shortened the cycle and reduced cost but lacked sufficient load-carrying capacity for service. Consequently, a GD strategy was adopted to reconcile manufacturability with mechanical performance [Figure 3(c)]. GD was performed in Autodesk Fusion 360 (GD workspace) to obtain a material-efficient internal architecture while preserving the functional geometry of the eccentric gear and ensuring feasibility for BMD. The GD study objective was set to minimize mass and material usage within a predefined design envelope while enforcing explicit geometric constraints (Figure 4). Preserved regions were defined to lock all fit-critical and load-transfer interfaces, including (i) the gear teeth along the pitch circle, (ii) the bore/shaft interface, (iii) the keyway and (iv) mating faces, to maintain assembly compatibility, torque transmission, contact integrity and structural performance. Obstacle regions were defined around required clearances to prevent material growth into assembly-restricted volumes. The starting volume was defined as the allowable internal design space and was intentionally biased toward efficient load-path formation by introducing three concentric circular patterns centered on the gear axis, reducing initial stock and encouraging truss-like solutions, as shown in Figure 4.

To emulate torque transfer, the applied tooth load was derived from the pump operating $T=P/ω$⁠, conditions. For a power of 0.25 kW at 1500 RPM, the transmitted torque was calculated using, yielding 1.59 Nm. Using the confirmed pitch circle diameter of 80 mm, the tangential force was obtained from $T=Ft(d/2)$⁠, resulting in a nominal transmitted load of 39.75 N. A factor of safety of 1.5 was applied to account for peak operating conditions and transient load amplification during gear meshing, leading to a conservative design load of approximately 60 N per tooth. This load was applied as a uniformly distributed normal force along the active tooth profile. To reproduce service loading conditions during GD solution generation and candidate selection, the gear was constrained at the inner bore (all translational degrees of freedom fixed to represent shaft mounting) [Figure 4(a)].

The GD solver generated multiple candidate geometries under identical constraints, and only solutions reported by Fusion 360 as “completed/converged” (i.e. the solver reached its termination criterion and produced a final solution rather than an intermediate or failed run) were considered in the selection step. Any runs marked as failed, incomplete, or nonconverged were excluded. The remaining solutions were then filtered using an explicit two-stage selection logic: (i) BMD printability/manufacturability and (ii) load-bearing performance under the operating load case. GD outputs were selected only if they were manufacturable using BMD. In this context, manufacturability means the geometry must remain a single, fully connected part and provide open, accessible pathways that allow binder removal during debinding. In addition, the selected GD outputs respected BMD-printable feature sizes, avoiding extremely thin walls or very small clearances that reduce print reliability. Among the BMD-feasible candidates, designs were then ranked using processing-efficiency criteria, prioritizing those that reduced Desktop Metal Fabricate™-predicted debinding–sintering cycle time and material consumption, while preserving all required interfaces. The shortlisted candidates were subsequently evaluated under the conservative 60 N per tooth load case using FEA verification, and only designs maintaining sufficient stiffness and strength at the tooth and hub regions were retained for final selection. The final GD design reported and fabricated in this work [Figure 4(b) and (c)] is the highest-ranked candidate under this screening workflow and was subsequently validated via FEA and experimental testing.

The selected GD variant delivered clear advantages relative to the fully solid design. Table 1 summarizes the final part weight, total consumed material, cost and fabrication times (printing, debinding and sintering) for the fully solid gear, 30% infill gear and GD gear. Metal consumption decreased from 1.16 kg to 0.842 kg (−27.4%), resulting in a 21.3% reduction in final part weight. Material costs dropped from $159.23 to $123.33 (−22.5%). The unrealistically predicted thermal cycle decreased from 708 h 56 min (655 h primary debinding) to 184 h 19 min (130 h primary debinding); a 74% reduction overall and an 80% reduction in primary debinding time. These improvements enable gear sintering within furnace limits while maintaining service-relevant stiffness at functional interfaces.

A linear elastic finite element simulation was performed for both the fully solid and generatively designed gear using ANSYS Workbench to evaluate their performance under the applied load. The solid gear 3D model was reproduced from the original gray cast iron gear through 3D scanning as discussed earlier in section 2.1, whereas the generative designed gear was obtained from the generative-design optimization study. Three different simulations were formed as per the real conditions with three different materials, where the gray cast iron is used for solid gear, while CoCr (produced via LPBF) and 4140 Steel (produced via BMD) were used for generative designed gears respectively. The elastic material properties were adopted from the tensile test data.

Furthermore, both the solid and generatively designed gears were discretized using three-dimensional tetrahedral elements [Figure 5(a) and (b)]. A mesh convergence study was performed to ensure numerical accuracy. The global element size was progressively refined, and an element size of 0.3 mm was selected as it yielded stable maximum von Mises stress values at the critical tooth root region (Figure 6). Further refinement below this size resulted in negligible variation in stress while significantly increasing computational time. Boundary conditions were defined to replicate service conditions. The bore region corresponding to the shaft interface was constrained in all translational degrees of freedom to represent rigid shaft mounting during operation (Figure 4). To emulate torque transfer, a uniformly distributed tangential load of 60 N per tooth was applied along the active tooth profile at the pitch circle diameter of 80 mm. This load was derived from the calculated operational torque of the pump assembly and includes a factor of safety of 1.5 to account for peak operating conditions (as calculated in Section 2.2). A full nonlinear tooth-to-tooth contact model was not implemented, as the objective of this study was comparative global stress evaluation rather than detailed contact stress analysis. Instead, the distributed loading approach was adopted, which is commonly used in preliminary gear strength assessments where structural response and material comparison are the primary focus. Similarly, the present FEA is limited to linear-elastic static loading and does not model fatigue damage accumulation; these aspects will be considered in future work. The simulations were performed with ASUSTEK ROG ZENITH Workstation equipped with an AMD Ryzen Threadripper 3970X 32-core processor and 256 GB of RAM. Depending on model complexity and resolution, each simulation required approximately 0.2–1.5 h.

This section used the inherent-strain based finite element solver within ANSYS Additive Print 2024 R1 to numerically predict build-process-induced distortion in the generatively designed gear. Build orientation in LPBF is a critical factor that governs the need and distribution of support structures and strongly influences final surface quality, dimensional accuracy and distortion behavior (Aiza et al., 2025). The objective of this stage was to conduct a comparative analysis of two representative orientations; a horizontal (0°) orientation and a 45° tilt relative to the build plate; to assess their impact on support requirements and predicted distortion.

The computational model was created from the STL geometry of the GD gear (volume 96,194.34 mm³; bounding box 82.02 × 84.07 × 68.56 mm³). The surface mesh, comprising 211,090 triangular facets, was discretized into a voxel-based Cartesian grid. A critical voxel size of 0.50 mm with a resolution factor of 5 was selected to balance geometric fidelity and computational efficiency. The inherent strain formulation was applied to capture layerwise thermal shrinkage, using anisotropic strain coefficients of 1.5 (x), 0.5 (y) and 1.0 (z). The material was modeled as CoCr using a J2 plasticity model with an elastic modulus of 205 GPa, Poisson’s ratio of 0.33, yield strength of 640 MPa and a kinematic hardening coefficient of 0.0148. The solver employed Dynamic Load Stepping to improve convergence during layerwise strain application. Processing parameters; laser power (160 W), scan speed (750 mm/s), hatch spacing (80 µm) and layer thickness (30 µm); corresponded to a volumetric energy density of 88 J/mm³, a window previously reported to yield near-full-density CoCr parts in LPBF studies (Liu et al., 2023). Post-processing of simulation results was performed using ParaView for visualization (Figure 7).

The predicted distortion field [Figure 7(a) and (b)] revealed a maximum displacement magnitude of approximately 0.38–0.40 mm in the horizontal (0°) orientation, primarily localized at the unsupported tooth tips and lower support interface regions. Distortion gradients were concentrated along the overhanging tooth geometry and at the transition between the web structure and the outer rim. In contrast, the 45° tilted orientation reduced the peak predicted displacement to approximately 0.25–0.32 mm, representing an estimated 16%–38% reduction in maximum distortion. The distortion pattern in the tilted configuration was more uniformly distributed, with lower stress concentration at critical tooth flanks and reduced displacement at the bore region.

This improvement is attributed to the altered thermal and mechanical boundary conditions introduced by the inclined build strategy. The horizontal configuration required extensive support beneath the gear teeth, increasing constraint stiffness and promoting localized residual stress accumulation at the support–part interface. By contrast, the 45° orientation reduced both the number and contact area of support structures, allowing more gradual heat dissipation and minimizing constraint-induced distortion. Such behavior aligns with previous findings (Aiza et al., 2025), which report that tilted orientations promote more uniform thermal gradients and reduced distortion in geometrically complex regions. Based on the combined reduction in peak displacement and improved deformation uniformity, the 45° orientation was selected for fabrication.

To establish quantitative and spatial validation of the distortion prediction, the fabricated LPBF-CoCr gear (produced in the selected 45° orientation) was subjected to detailed scan-to-CAD dimensional evaluation as described in Section 2.5.3. The deviation map generated in Geomagic Control X revealed that the majority of the surface remained within a narrow tolerance band (±0.05 mm), with a best-fit surface standard deviation of 0.0826 mm. Importantly, the regions exhibiting the highest measured deviations corresponded to the same geometric zones identified in the simulation as distortion-sensitive, namely the outer tooth perimeter and the web-rim transition area. In contrast, the bore region, which showed minimal predicted displacement in the 45° configuration, also demonstrated tight dimensional control within ISO 286 H7 tolerance limits. The reported peak displacements for each orientation were extracted from the ANSYS Additive Print displacement results. It is noted that inherent-strain simulation represents build-stage deformation, whereas scan-to-CAD deviation reflects the final geometry after support removal and stress relaxation; therefore, validation is based on quantitative deviation statistics and spatial trend agreement rather than direct one-to-one matching of a single maximum value. This spatial correspondence between predicted deformation patterns and measured geometric deviation confirms the reliability of the inherent-strain simulation for orientation selection and distortion risk assessment in complex gear geometries.

To enable a comparative evaluation of process-dependent effects on dimensional fidelity, build quality and post-processing effort, the generative designed gear was fabricated using two metal AM techniques: LPBF and BMD. This dual approach also facilitated material-based comparisons, with CoCr used for LPBF and 4140 steel for BMD.

LPBF fabrication was performed on a 3D Systems DMP Flex 200 under an argon atmosphere, using a 500 W fiber laser. The feedstock was a gas-atomized CoCr alloy with a particle size distribution of 15–45 μm. The process parameters, including a layer thickness of 30 μm, laser power of 160 W, scan speed of 750 mm/s and hatch spacing of 80 µm, were used as input to mirror the conditions used in the ANSYS Additive Print simulations and to align with literature-reported processing windows for near-full-density CoCr (Liu et al., 2023). Based on the simulation results for different orientations, the 45° tilt was selected for LPBF production of the gear (Figure 8). This configuration demonstrated superior outcomes, requiring fewer supports and reducing predicted distortion compared to the horizontal build. The trade-off was an increased build height, which led to a longer fabrication time (18 h 7 min) and a greater number of layers (2406). By contrast, the horizontal orientation would have reduced the build time (15 h 25 min, 1594 layers). However, the gear’s geometry in this orientation necessitates a continuous support body beneath the entire base, which significantly increase the support-part contact area. This not only raised the post-processing effort for support removal but also elevated the risk of surface damage and distortion at the bottom features. Such effects have been reported by Dallago et al. (2019), who observed that horizontal builds often produce struts thicker than designed due to localized overheating and poor thermal conduction in the surrounding loose powder. The printed part was removed from the substrate via wire EDM. Following support removal, the part underwent shot blasting with aluminum oxide media under controlled pressure to remove residual powder and homogenize the surface finish. This step was critical to improving the surface condition of the CoCr part, which exhibited the characteristic LPBF-induced roughness from partially melted particles and layer stratification. No heat treatment or hot isostatic pressing (HIP) was applied, as the aim was to test the as-built performance.

To extend the comparative scope beyond laser-based powder bed fusion, the GD gear was also fabricated using BMD. BMD belongs to a class of sintering-based metal AM processes in which premixed metal-binder feedstock is deposited to form a “green” part that is subsequently debindered and sintered to yield a fully metallic component. Unlike LPBF, which densifies via localized melting and rapid solidification, BMD relies on binder removal followed by solid-state diffusion and phase transformation during sintering, leading to distinct microstructural and density-evolution pathways (Basak et al., 2024). Parts were produced on the Desktop Metal Studio System 2, which uses a two-step route in this generation: (i) printing of a bound-metal green part and (ii) a single furnace cycle that combines thermal debinding and sintering. The green parts were scaled 18% linearly to compensate for sintering contraction. Figure 9 shows the full build layout for 4140 steel, including full and sectional view of GD gears.

During printing, material was extruded through a heated nozzle onto a temperature-controlled stage using the manufacturer-recommended parameters for 4140 steel (nozzle temperature 165°C, build plate temperature 50°C, layer height 0.15 mm, printing speed 30 mm/s, nozzle diameter 0.4 mm, nominal 100% infill; machine-specific settings as configured). Thermal debinding and sintering were performed in the Studio furnace in a single cycle under a pure argon atmosphere (manufacturer-specified profile for 4140; peak temperature in the high-temperature austenitic range; total cycle on the order of tens of hours) to promote densification and metallurgical bonding. Separable support interfaces (ceramic release layers) generated during printing were manually removed after sintering; their placement is critical for shape stability during shrinkage while enabling low-force removal post-furnace. All printed parts underwent shot blasting under the same conditions used for LPBF parts to remove surface residue and standardize roughness prior to evaluation. No post-sintering heat treatment was applied; all reported properties reflect the as-sintered condition.

The mechanical properties of additively manufactured components are critical to validating their performance for industrial applications. In this study, tensile and hardness tests were conducted to assess the mechanical behavior of parts produced using LPBF and BMD with two materials: CoCr and 4140 steel, respectively. These properties were compared against the reference gray cast iron part, which served as the baseline for evaluating the feasibility of AM technologies.

Tensile tests were performed on dog-bone-shaped specimens prepared from the two AM materials and the cast iron reference part. Testing was carried out according to ASTM E8/E8M standards using a universal testing machine. The tensile specimens (Figure 10) had a gauge length (G) of 10 mm, a width (W) of 2.4 mm and a thickness (T) of 2.4 mm. The radius of the fillet (R) was 2 mm, with an overall length (L) of 40 mm and a reduced parallel section (A) length of 12.8 mm. The grip section (B) was 12 mm in length, with a width (C) of 4 mm. These dimensions were scaled down proportionally from the standard geometry to accommodate the geometric constraints of the original gray cast iron gear. The additively manufactured tensile specimens (BMD-4140 and LPBF-CoCr) were fabricated with identical dimensions to enable direct comparison with the cast iron reference and to ensure consistency in testing.

The tests were conducted using a universal testing machine equipped with precise load and displacement measurement capabilities. The strain rate was set to 0.015 mm/mm/min, resulting in a crosshead speed of 0.15 mm/min, in compliance with the ASTM standards for tensile testing. All tests were performed in ambient laboratory conditions. Load and elongation were recorded continuously throughout the test until the specimen fracture. After failure, fracture surfaces were examined using scanning electron microscopy (SEM) to document failure mode and to identify any process-related defects that may have influenced mechanical performance.

Complying with the ASTM E92 standard, Vickers microhardness testing was performed to assess the surface mechanical properties of the three investigated materials: 4140 steel fabricated via BMD, gray cast iron (reference sample) and CoCr produced using LPBF. Testing was conducted on the polished external surface of each specimen using a Pace Technologies MHV-2000 microhardness tester. A load of 4.903 N was applied with a dwell time of 10 s. Prior to testing, surfaces were prepared to a 1 µm diamond finish to minimize roughness effects, and the instrument was verified on a certified reference block. For each material, three indentations were made in representative regions of the sample surface to ensure repeatability, and the average hardness value was calculated and reported for further analysis.

Dimensional accuracy was evaluated using high-resolution 3D scanning (see Section 2.1) and inspection within Geomagic Control X, with the original CAD serving as the reference model. The digital inspection protocol began with a global best-fit alignment of the scan to CAD, followed by a datum-based refinement using the bore axis as the primary datum and the keyway plane as a secondary constraint, after which a color-coded deviation map (tolerance band ± 0.05 mm) was generated to visualize geometric departures at fit-critical features (Figure 11). To quantify accuracy, feature-level measurements were extracted from both CAD and scan models. This included upper- and lower-hub diameters and heights, overall gear height and gear-specific features such as tooth thickness at the pitch circle and circular pitch; in addition, keyway width and keyway length were measured due to their criticality for shaft coupling in pump assemblies. Features were selected for their mechanical relevance and functional role in torque transfer, alignment and assembly interchangeability.

Feature tolerances and inspection thresholds were established from relevant ISO standards (Table 2). For gear features (circular pitch, tooth thickness), tolerancing followed ISO 1328–1:2013 [33], as a benchmark for flank-related deviations (pitch/profile) and quality grades. For general linear dimensions (hub/shaft diameters, gear height), limits referenced ISO 2768–1:1989 [34] (general tolerances) and ISO 286–1 / 2:2010 [35,36] (fits and tolerances for bores/shafts, including keyway geometry; see Figure 12). Although ISO 1328–1 targets machined gears, it provides a useful reference framework for assessing AM-achievable accuracy. This inspection strategy enabled a standardized, process-agnostic comparison of dimensional performance between LPBF and BMD, clarifying their ability to meet industrial tolerance targets and their suitability for functional gear production in energy-sector pump systems.

The mechanical properties, including ultimate tensile strength and elongation at fracture for tensile test samples of 4140 steel produced by BMD, a CoCr alloy produced by LPBF and conventionally cast gray iron, are presented in Figure 13(a) and (b), respectively. The data reveal clear differences in mechanical performance among the three investigated materials reflecting both their intrinsic alloy characteristics and process-dependent microstructures. The LPBF-fabricated CoCr alloy exhibited the highest ultimate tensile strength (UTS), reaching 1247 MPa, which is nearly three times that of cast iron (430 MPa) and more than double that of BMD 4140 steel (599 MPa). The high strength of the LPBF alloy can be attributed to rapid solidification during laser processing, which produces a fine cellular microstructure, combined with solid solution strengthening in the CoCr matrix. The SEM images (Figure 14, middle row) confirm this, showing undulating fracture surfaces, tear ridges, cleavage facets and micro-cracks, features characteristic of a mixed ductile-brittle response. The presence of striations suggests localized plastic deformation prior to fracture, consistent with the moderate elongation observed. Elongation at failure [Figure 13(b)] was highest for BMD 4140 steel (0.35 mm), exceeding that of both LPBF CoCr (0.26 mm) and cast iron (0.19 mm). The relatively balanced performance of BMD 4140 reflects the homogeneous microstructure achieved after binder removal and solid-state sintering, where diffusion bonding reduces porosity but leaves some residual sintering pores.

SEM images of BMD 4140 (Figure 14, top row) show layered extrusion traces, cleavage steps and sintering-induced voids, explaining the intermediate tensile strength but relatively high ductility. This ductility advantage is industrially relevant, as it reduces the risk of catastrophic brittle fracture under fluctuating service stresses. The gray cast iron reference demonstrated the lowest UTS (430 MPa) and the lowest elongation (0.19 mm), confirming its inherently brittle behavior. SEM analysis (Figure 14, bottom row) revealed cleavage-dominated fracture with graphite flake debonding, a well-known microstructural feature that acts as a crack initiator and propagator, thereby limiting tensile ductility. These results highlight that AM technologies can significantly outperform conventional casting in both strength and ductility. LPBF CoCr delivered the highest strength, suitable for applications where load capacity and wear resistance are critical, while BMD 4140 offered superior ductility and reasonable strength, making it attractive for applications requiring tolerance to impact or cyclic loading. In contrast, cast iron, although traditionally used in pump gears, exhibited the weakest mechanical response, underscoring the potential of AM to provide functionally superior replacements for legacy components in energy-sector equipment.

Microhardness testing was conducted to quantitatively assess the mechanical performance of the three materials investigated, providing insight into the combined influence of feedstock composition and manufacturing process on localized strength. The results, summarized in Figure 15, reveal pronounced differences in hardness across the three material-processing combinations. The LPBF-CoCr specimens recorded the highest hardness values, averaging 443 HV, with individual measurements ranging between 400 and 470 HV. This superior hardness is attributed to the rapid solidification and high thermal gradients in LPBF, which promote a fine-grained microstructure reinforced by solid solution strengthening. Such characteristics not only enhance wear resistance but also align with literature reports of LPBF CoCr alloys used in aerospace and biomedical applications. In contrast, the BMD-processed 4140 steel exhibited the lowest hardness at 164 HV, reflecting the intrinsic limitations of sinter-based processes. The reduced hardness can be traced to the porosity introduced during debinding and sintering. While this value is below that of wrought or quenched-and-tempered 4140 steel, it remains consistent with as-sintered AM steels, confirming the need for post-sintering heat treatments to unlock the full performance potential of BMD alloys. The cast iron reference demonstrated an intermediate hardness of 368 HV, consistent with its well-known pearlitic-ferritic matrix containing dispersed graphite flakes. This structure provides significant compressive and wear resistance but inherently reduces tensile ductility, as confirmed in the mechanical tests.

The structural response of the original cast iron gear and the generatively designed (GD) gears was investigated through finite element simulations to assess their mechanical performance under operational loading. The analyses focused on maximum deformation and von Mises stress as key indicators for comparative evaluation across the three materials and manufacturing routes. The finite element simulations clearly demonstrate that the GD-CoCr gear achieves the lowest deformation (0.002159 mm) and stress (11.9 MPa) among the three investigated materials, outperforming both the original cast iron gear and the GD-4140 steel gear. Importantly, all the maximum stress values observed from the simulation results (11.9 MPa for CoCr, 12.8 MPa for 4140 steel and 13.64 MPa for cast iron) are significantly lower than the respective yield strengths, indicating that all designs are safe under the simulated loading conditions.

In terms of deformation, the original cast iron gear showed a maximum deformation of 0.00254 mm, while the GD CoCr part exhibited the lowest deformation at 0.002159 mm, reflecting its superior stiffness. In comparison, the GD-4140 steel gear displayed the highest deformation (0.002631 mm), which can be attributed to its lower yield strength relative to CoCr [Figure 16(a)]. The simulation result for the deformation behavior of gears is also presented in Figure 17(a)–(c). Where the different views of original and generative designed gears are presented for comparison. These results indicate that the generative-designed gear made of CoCr maintains greater dimensional stability than both the cast iron gear and the GD-4140 gear, making it the most effective option for current applications requiring minimal deflection under load. A similar trend was observed in stress results [Figure 16(b)]. The original cast iron gear experienced a maximum von Mises stress of 13.64 MPa, while the GD-CoCr gear exhibited a reduced stress of 11.9 MPa, highlighting the effectiveness of the optimized geometry in redistributing loads. The GD-4140 gear recorded an intermediate stress of 12.8 MPa, lower than cast iron but higher than CoCr. Stress contour plots drawn at a similar scale [Figure 18(a)–(c)] confirmed that the generative-designed models experienced more uniform stress distribution, with fewer localized concentrations compared to the original solid cast design. This demonstrates the ability of GD to enhance load distribution, particularly when combined with high-performance AM materials.

The superior performance of the GD-CoCr gear is supported by the experimental results (Figure 13), where the LPBF-processed CoCr alloy exhibited the highest ultimate tensile strength (1247 MPa) alongside appreciable ductility. The combination of optimized GD and the high-performance microstructure achieved through rapid solidification in LPBF, including solid solution strengthening, allows the GD-CoCr gear to maintain low stress concentrations and excellent load-bearing capacity. The smoother stress distribution observed in the CoCr simulation aligns with its experimentally confirmed ability to sustain high stress without localized yielding or premature failure.

The GD-4140 steel gear exhibited the highest deformation (0.002631 mm) and intermediate stress (12.8 MPa), consistent with its experimental mechanical behavior. BMD-processed 4140 steel has a yield strength of 415 MPa and UTS of 599 MPa, which is lower than CoCr but higher than cast iron. While the maximum stress is far below its yield limit, the lower intrinsic strength relative to CoCr leads to comparatively higher deformations under similar loading. The simulation also shows no sharp stress concentrations, which matches the mechanical response observed experimentally, indicative of uniform microstructure and ductile behavior. In contrast, the original cast iron gear demonstrated both higher stress (13.64 MPa) and relatively high deformation (0.00254 mm) compared to the GD-CoCr design. Despite the low stresses being well below its yield strength (276 MPa), the inherently brittle nature of cast iron, with UTS of 430 MPa and elongation at failure 0.19, makes it more susceptible to localized failure.

Taken together, the close agreement between experimental and simulation results confirms that all designs are structurally safe under the loads considered. However, when comparing materials and manufacturing technologies, there are notable variations in mechanical performance. The original cast iron gear, provided by industry and produced via conventional casting, exhibited limited strength and ductility, reflecting its intrinsic brittleness. By reproducing and redesigning this gear using AM techniques, LPBF for CoCr and BMD for 4140 steel, we were able not only to select higher performance materials but also to optimize the internal geometry through GD. This approach allowed for smoother stress distribution, reduced deformation and improved load-bearing capacity compared to the original cast gear. The GD-CoCr gear, in particular, benefits from both the inherent material strength of LPBF-processed CoCr and the optimized internal structure, demonstrating that AM combined with GD offers a clear advantage over traditional casting methods in producing safer, lighter and more efficient gears.

A detailed dimensional inspection was carried out to assess how closely the additively manufactured gears replicated the nominal CAD geometry. Ten critical features were evaluated for both LPBF-CoCr and BMD-4140 components, covering diameters, axial heights, gear teeth geometry and keyway dimensions, as presented in Table 3. The evaluation criteria were established in accordance with ISO standards: ISO 1328–1:2013 for gear-specific features, ISO 2768–1:1989 for general tolerances on linear and angular dimensions and ISO 286–2:2010 for fits and mating features such as bores and keyways. The LPBF-CoCr gear demonstrated higher dimensional fidelity compared to the BMD-4140 gear, with a best-fit surface standard deviation of 0.0826 mm (Figure 19). Most features remained within their prescribed ISO tolerance bands. However, slight undersizing trends were observed, with mean deviations of approximately −0.18 mm for diametric features and −0.12 mm for axial heights. These deviations are consistent with thermal contraction during rapid solidification and repeated layer re-melting, phenomena commonly reported in LPBF processes.

As illustrated in Figure 20, a critical observation was the bore diameter deviation of −0.0214 mm, which falls exactly at the lower limit of the ISO 286 H7 tolerance band (±0.021 mm). This deviation is likely linked to the 45° build orientation, where thin-walled cylindrical features are periodically exposed to recoater blade interaction. Such repeated lateral contact may accumulate small plastic deformations, especially in unsupported regions. Despite this, the LPBF-CoCr gear maintained high accuracy across complex features, including tooth thickness (+0.038 mm) and circular pitch (+0.079 mm), both of which remain within ISO 1328–1 quality grade Q8-Q10 tolerances. These findings confirm that LPBF can reliably reproduce functionally critical features with tight tolerances, making it highly suitable for gears requiring precise meshing and engagement.

In contrast, the BMD-4140 gear exhibited greater variability, with a best-fit surface deviation of 0.1268 mm. Some features showed good compliance, such as gear height (−0.058 mm) and upper hub height (+0.033 mm), both within ISO 2768-m tolerances. Similarly, the circular pitch deviation (−0.018 mm) was within the acceptable ISO 1328–1 range. However, a significant oversizing of the bore diameter (+0.109 mm) was recorded, far exceeding the H7 limit. This error suggests anisotropic shrinkage and radial expansion during sintering, a known limitation of sinter-based processes where nonuniform densification affects accuracy. Such deviations are difficult to correct without robust calibration and compensation models. The increased bore deviation observed in the BMD 4140 gear (+0.109 mm vs nominal 19 mm) is consistent with process-driven radial expansion during the green-to-sinter cycle rather than scanner or CAD errors alone. Contributing mechanisms include: (1) heterogeneous packing density in the green part leading to nonuniform shrinkage, (2) differential sintering rates between thick and thin sections, where thicker regions densify slower and can produce local tensile stress and radial expansion, (3) thermal gradients within the furnace and fixture-induced constraint from friction against supports or ceramic release layers and (4) potential overcompensation from a uniform linear scaling of the green geometry that does not capture local anisotropy. These phenomena are well-documented in the BMD literature (Gabilondo et al., 2022; Di Pompeo et al., 2023; Basak et al., 2024). Overall, the results highlight a clear contrast between the two AM technologies. While BMD provides stable global shape retention, its dimensional precision in fine features can be less predictable, particularly for mating geometries. In contrast, LPBF-CoCr achieved tighter control over both global and local features in this case study, making it a strong option when high dimensional accuracy, fit and functional meshing are critical (Figures 19–20). Conversely, BMD remains a practical alternative when cost efficiency, scalability and adequate toughness are prioritized and fine-feature tolerance demands are less stringent.

A gear contact check in accordance with API Standard 677 was carried out to evaluate the meshing behavior and alignment before the generatively designed additively manufactured gear was deemed fully aligned and operationally ready [Figure 21(a)]. The objective of this method is to make sure there is no misalignment or inappropriate contact, and observe how the gear teeth mesh under regulated, low-stress circumstances. This stage is essential to guarantee that the final alignment and installation are accurate, that the gears operate smoothly, and to prevent problems like noise, vibration, uneven wear, or gear failure in the future. A thin marking compound was applied to selected teeth at multiple circumferential locations, the assembly was rotated under light resistance, and the contact imprint was examined on both flanks. The check was conducted on-site together with the operating company’s maintenance engineer as part of standard pre-start commissioning practice for gear-driven assemblies. The pattern was centered on the tooth face with continuous contact and no edge loading or localized spotting, indicating correct axial and radial alignment [Figure 21(b)]. This allowed us to inspect the alignment and contact pattern without damaging the teeth or inducing deformation. Following the contact verification, the gear was installed in the chemical pump [Figure 21(c)]. After installation, a run was performed under normal operating conditions to confirm functional readiness. The unit was started and brought to steady operation, then observed during a brief initial run period (45–60 min) at nominal operating speed (approximately 1450 rpm). During the commissioning run, the drivetrain exhibited smooth operation with no abnormal noise or vibration, and rotation was free of binding. The gear functioned as intended under operational conditions, confirming that the additively manufactured, generatively designed component meets the functional requirements of API 677 contact acceptance and is suitable for in-service deployment.

To ensure a structured and application-driven comparison between LPBF and BMD technologies, the evaluation in this study was conducted across explicitly defined decision metrics relevant to critical energy-sector spare parts. These metrics include: (i) mechanical performance (ultimate tensile strength, elongation, hardness and FEA-derived deformation), (ii) dimensional accuracy (scan-to-CAD deviation and ISO tolerance compliance), (iii) economic viability (material consumption, fabrication cost and overall process duration), (iv) lead time and workflow complexity (printing, debinding, sintering and post-processing burden), (v) failure risk profile (brittleness versus ductility behavior under service loads) and (vi) process reliability and scalability (distortion sensitivity, shrinkage control and reproducibility). This structured framework enables objective process selection based on functional criticality rather than descriptive observation.

Evaluated against the defined decision metrics, the comparative analysis between LPBF-CoCr and BMD-4140 demonstrates that while both technologies can successfully reproduce complex industrial components, their performance is defined by distinct process-structure-property relationships. LPBF, owing to its rapid melting and solidification cycles, produces a refined microstructure that imparts superior ultimate tensile strength (1247 MPa) and hardness (443 HV). These results are further corroborated by the finite element simulations, where the LPBF-CoCr gear exhibited the lowest deformation and von Mises stress among all tested materials, aligning with its experimentally validated mechanical robustness. The fine solidification microstructure not only contributes to strength but also ensures dimensional stability, as confirmed by the low surface deviation of 0.0826 mm relative to the CAD reference. Within the defined framework, LPBF therefore ranks highest in mechanical performance and dimensional accuracy, making it particularly suitable for safety-critical applications where both strength and geometric fidelity are nonnegotiable.

BMD, on the other hand, follows a fundamentally different consolidation pathway governed by binder removal and high-temperature solid-state sintering. The resulting 4140 parts, while exhibiting moderate hardness (164 HV) and strength (599 MPa), showed higher elongation at fracture (0.35 mm) compared to both LPBF-CoCr and cast iron, indicating a favorable toughness profile. This ductility, coupled with the relatively uniform stress distribution observed in simulations, suggests that BMD-processed steels can sustain mechanical loading without catastrophic failure, even if their strength is lower. However, the dimensional evaluation revealed higher variability (0.1268 mm standard deviation), particularly in critical features like bore diameter. These findings emphasize that while BMD is highly cost-effective and operationally simpler, additional calibration or compensation strategies are required to reliably achieve the precision necessary for tightly tolerance industrial parts. Within the evaluation framework, BMD demonstrates strong economic viability and process scalability, but requires careful shrinkage compensation and calibration strategies to consistently achieve tight tolerances.

An important correlation emerging from this study is that the differences in performance between LPBF and BMD are not solely due to intrinsic material properties, but rather the processing routes that govern defect formation, densification behavior and microstructural evolution. LPBF benefits from complete melting and rapid cooling, but this comes at the expense of residual stress, sensitivity to orientation and higher processing costs. In contrast, BMD avoids the complexity of powder bed fusion but introduces anisotropic shrinkage during sintering, requiring careful design compensation and robust post-processing to ensure fidelity. The integration of GD across both processes proved essential in optimizing material efficiency and minimizing distortion, highlighting the importance of coupling advanced design strategies with AM technologies to achieve practical, industrially relevant outcomes.

Consistent with the defined evaluation framework, Figure 22 presents a multi-criteria radar constructed across five decision axes: mechanical performance, dimensional accuracy, process reliability and scalability, economic viability and industrial integration potential. The relative influence of each axis was governed by functional criticality. Within this framework, LPBF scored higher in mechanical performance and dimensional accuracy, whereas BMD scored higher in process reliability/scalability and economic viability; both show strong industrial integration potential as evidenced by the successful field deployment. This multi-criteria synthesis underscores the necessity of application-specific decision-making: LPBF may be preferred when tight tolerances and high strength/hardness (with associated wear resistance) are required for fit- and function-critical components, whereas BMD may be preferred for spare parts where turnaround time, cost efficiency and adequate toughness/ductility are prioritized over maximum strength.

From a broader perspective, the findings of this study provide more than a direct comparison of two AM technologies; they establish a replicable, multi-criteria methodology for evaluating process suitability across industrial applications. By combining reverse engineering, GD, finite element simulations, dimensional inspection and real-world validation, this work presents a decision-support framework for engineers tasked with replacing or upgrading critical components in the energy sector. Looking forward, future research should expand on these findings by incorporating long-term durability assessments, including fatigue life, corrosion resistance and in-service monitoring, as well as exploring hybrid strategies where different AM routes (and conventional processes) can be used in a complementary manner. Such pathways may enable the tailoring of performance and cost, further accelerating the adoption of AM in energy and other high-stakes industrial domains. While this study demonstrates functional feasibility through dimensional inspection, static structural simulation, contact-pattern verification and short-term field operation, it does not include fatigue-life prediction or detailed tooth-to-tooth contact (Hertzian) stress modeling, and it does not quantify long-term durability mechanisms such as fatigue life, tooth-flank wear or time-dependent contact-stress evolution under realistic load spectra. Accordingly, the present results should be interpreted as workflow and performance at commissioning validation rather than a lifetime qualification; extended service monitoring and dedicated fatigue and wear testing will be addressed in future work.

Fatima Ghassan Alabtah: Conceptualization, methodology, validation, formal analysis, investigation, data curation, writing – original draft, writing – review & editing, visualization; Manel Chihi: Methodology, validation, formal analysis, modeling and simulation, writing – original draft, writing – review & editing; Abdalla Mohammed: Modeling and simulation, writing – review & editing; Numan Khan: Modeling and simulation, validation, formal analysis, writing – review & editing; Yasser Al Hamidi: Methodology, validation, resources, writing – review & editing; Mohammad Albakri: Methodology, validation, resources, writing – review & editing; Marwan Khraisheh: Conceptualization, writing – review & editing, supervision, resources, project management, funding acquisition.