Lightweight sandwich structures fabricated by fused filament fabrication (FFF) require geometric designs that balance mechanical performance and material use. Polylactic acid (PLA) is commonly used for FFF due to its processability and specific mechanical properties. In sandwich panels under bending, face sheets resist normal stresses while the cellular core transfers shear and maintains separation. Changes to core and face-sheet dimensions affect both load-bearing behaviour and material requirement. This study experimentally investigates four geometric variables—honeycomb cell size (HCS), honeycomb cell-wall thickness (HCT), top face-sheet thickness (TFS), and bottom face-sheet thickness (BFS)—and aims to answer three design questions: which parameters predominantly control flexural strength and mass; whether a single parameter set can satisfy both competing responses; and whether a contribution-weighted combination of response-specific settings yields a practical strength–mass compromise.
The objectives were to (i) quantify the effects of HCS, HCT, TFS, and BFS on measured flexural strength and panel mass; (ii) determine the statistical contribution and significance of each parameter; (iii) identify and reconcile conflicting factor settings associated with the two responses; and (iv) experimentally evaluate the resulting compromise configuration. The principal contribution is the integration of experimentally obtained factor-level trends and response-specific contribution ratios to select a single strength–mass compromise for experimental evaluation.
All specimens were printed from commercial PLA filament (1.75 mm diameter, manufacturer recorded as eSUN) and used as received without additional drying or reheating. Printing used a PRATHAM 5.0 FFF machine with a 0.4 mm nozzle. The fixed printing parameters for all specimens were: extrusion temperature 210 °C, nominal infill density 100%, layer thickness 0.2 mm, print speed 30 mm/s, bed temperature 55 °C, and cooling fan switched off. Wall/perimeter count and extrusion width were not recorded in the available process records; therefore, the exact slicer perimeter/extrusion settings cannot be reconstructed from the accessible documentation. All panels were printed in a flat orientation with the bottom face on the build plate; the longitudinal panel direction was aligned with the printer X-axis and honeycomb cell walls built vertically along the Z-axis. These fabrication conditions were held constant so that the experimental analysis focused on the selected geometric variables.
Specimens were modelled in Autodesk Fusion 360 and sliced using Simplify3D to generate G-code. Four geometric parameters were investigated: HCS (two levels), and HCT, TFS, BFS (three levels each). A conventional hexagonal honeycomb topology was retained for all specimens so that analysis isolated dimensional effects rather than topology. Because HCS was varied while keeping cellular count and topology consistent, changing HCS altered overall specimen length and the support span used during testing; thus the HCS contribution should be interpreted as the response of the complete geometric configuration rather than a pure single-cell-size effect.
A mixed-level Taguchi L18 orthogonal array organized the experimental programme, producing 18 distinct geometric configurations. One specimen was printed and tested for each L18 configuration; the reported measurements therefore represent single experimental observations rather than replicated means. Table-level details of factor levels and resulting specimen dimensions were recorded in the study's supporting information (S1). One additional specimen was fabricated and tested as a confirmation (validation) of the selected compromise configuration.
Three-point bending tests followed the test arrangement described in ASTM D7249/D7249M-20. A universal testing machine with a 10 kN load cell was used. To provide a common comparative condition across specimens of different heights, the support span was adjusted to maintain a nominal span-to-height ratio of approximately 3:1. Loading was applied at a constant crosshead displacement rate of 2 mm/min using a cylindrical loading nose with a 30 mm diameter. The maximum recorded load from each test was used to compute the reported nominal flexural-strength response of the complete sandwich specimen under the specified bending arrangement. Because the span-to-height ratio and specimen geometry were study-specific comparative conditions, the reported flexural-strength values are interpreted as nominal responses of the assembled sandwich configurations rather than as intrinsic material flexural properties.
Post-test photographs documented visible deformation, but the available experimental record did not include a systematic failure-mode classification or detailed close-up measurements of damaged regions. The original calculation sheet containing individual maximum-load values and the exact flexural-strength computation expression was not available among the accessible records; therefore the raw maximum-load numbers and the explicit formula used in the original calculations are not reported here.
Panel mass was measured directly for each specimen, and a flexural-strength-to-mass ratio was computed from the measured responses. The data underlying the reported findings are provided in the study's Supporting Information files S1, which contain experimental results for all 18 configurations and the confirmation configuration together with parameter settings and response values.
Analysis of means (ANOM) was used to identify the factor levels favoured independently by maximum nominal flexural strength and by minimum panel mass. Analysis of variance (ANOVA) quantified the relative contributions of the four geometric parameters to each response within the modelled design space. For flexural strength, HCS and HCT were the dominant contributors; for panel mass, HCT and HCS dominated the modelled variation.
Because factor-level settings that favour maximum strength differ from those that minimize mass, the authors combined the ANOM-identified response-specific factor levels with normalized ANOVA contributions in a contribution-weighted multi-response selection procedure. This procedure produced a single compromise parameter set intended to balance the competing objectives of strength retention and mass reduction within the investigated design space. The method produced a study-specific candidate configuration for experimental validation rather than a general optimization across all possible parameters.
From the ANOVA performed on the measured responses, the modelled contributions were reported as follows: for flexural strength, HCS accounted for 45.9% of modelled variation and HCT for 34.1%; for panel mass, HCT accounted for 50.9% of modelled variation and HCS for 34.0%. Face-sheet thicknesses (TFS and BFS) had smaller relative contributions within the investigated levels. These results indicate that core-related variables (HCS and HCT) dominated the modelled sensitivity of both measured responses in the chosen design space.
The contribution-weighted combination of ANOM factor levels and normalized ANOVA contributions yielded a compromise configuration with nominal CAD dimensions: HCS = 6 mm, HCT = 1.08 mm, TFS = 1.18 mm, and BFS = 1 mm. This configuration was fabricated and experimentally evaluated as the confirmation specimen.
In the validation experiment, the selected compromise configuration produced a measured nominal flexural strength of 13.21 MPa and a mass of 34.367 g. Compared with the highest-strength configuration among the L18 set, the selected compromise retained 90.54% of measured flexural strength while reducing panel mass by 50.29%. These comparative outcomes demonstrate that the contribution-weighted selection produced a practical strength–mass trade-off within the investigated parameter space.
The study demonstrates that within the investigated design space and fixed printing conditions, core geometry—represented by honeycomb cell size and cell-wall thickness—accounts for most modelled variation in both nominal flexural strength and panel mass. The factor settings that maximize strength were not identical to those that minimize mass; a contribution-weighted multi-response selection provided a transparent, study-specific route to choose a compromise design and validate it experimentally.
Limitations in the accessible experimental record include the use of single specimens per configuration (no replicated means), missing slicer extrusion-width and wall/perimeter count settings, absence of the original maximum-load calculation sheet, and lack of a systematic failure-mode classification. The authors report that all primary data underlying the findings are included in the Supporting Information S1 accompanying the article. No specific funding was received for the work, and the authors declared no competing interests.
Within these constraints, the contribution-weighted approach provided a practically useful method to reconcile competing objectives for lightweight FFF-printed honeycomb sandwich panels. The findings are specific to the tested geometry ranges, printing conditions, and the conventional hexagonal topology used in this study; extrapolation beyond these conditions would require additional experimental evaluation.