Experimental Investigations Related to Part Strength and Shape Realizability in Fused Filament Fabrication (FFF) Process

Abstract

Fused filament fabrication (FFF) is a very popular additive manufacturing technique that fabricates functional parts with complex geometrical features directly from 3D CAD models in a reasonably good time. In the last few decades, fused filament fabrication, also known as fused deposition modelling has been widely explored as one of the low-cost additive manufacturing (AM) processes. FFF is known for its cost-effectiveness, ease of use, and material versatility, making it a popular choice in industries such as aerospace, healthcare, automotive, and consumer goods. However, challenges such as anisotropic strength, dimensional accuracy, and surface roughness continue to remain areas of research and development. Hence, it becomes essential to thoroughly understand the inherent limitations of the FFF process and identify the key controllable process parameters that can be optimized to enhance the quality of the fabricated parts. In this study, polycarbonate (PC) has been chosen as the material of study owing to its superior mechanical properties, including high strength, impact resistance, and thermal stability, making it ideal for demanding applications in aerospace, automotive, and industrial manufacturing. The study focuses on improving the mechanical properties, dimensional accuracy, and surface quality of FFF-printed PC parts by investigating the influence of four key process parameters: layer thickness, extrusion temperature, printing speed, and extrusion width. As an application of FFF in tooling, the feasibility of fabricating EDM electrodes using rapid tooling techniques was explored. A parametric study has been carried out to quantify the mechanical properties of polycarbonate parts fabricated using the FFF process. Test specimens were produced following the applicable ASTM standards for tensile, flexural, compressive, shear, and impact testing. response surface methodology (RSM) and dimensional analysis were employed to model and analyze the interaction between input parameters and mechanical responses. The results indicated that the following parameters: an extrusion temperature of 270 °C, a layer thickness of 240 µm, a printing speed of 20 mm/s, and an extrusion width of 0.48 mm provides the optimum mechanical strength. Fractographic analysis using field emission scanning electron microscopy (FESEM) further validated these findings by revealing the microstructural features associated with each failure mode, including poor adhesion, filament pull-out, and crack propagation along layer lines. Dimensional accuracy and geometric accuracy serve as essential quality indicators in FFFprinted components, especially in engineering applications that demand strict tolerances. This phase of the study aimed to quantify and reduce the dimensions and geometric deviations of PC components manufactured using the FFF technique. To model the complex and nonlinear interactions between these variables and the observed deviations, three machine learning regression models were developed: linear regression (LReg), random forest regression (RFreg), and extreme gradient boosting regression (XGBReg). Among them, the XGBReg model demonstrated superior predictive accuracy in terms of MSE, RMSE, MAE and R2 effectively capturing nonlinear relationships and interactions between input variables. Results confirmed that lower layer thickness, lower printing speed, and narrower extrusion width in combination with higher extrusion temperature contributed to enhanced dimensional accuracy by improving deposition precision and reducing thermal deformation during the build process. Surface quality, which includes both roughness and waviness, is critical to the functional performance, aesthetics, and post-processing needs of FFF components. The effect of the selected process parameters on surface roughness and waviness was thoroughly investigated. Two modeling approaches were employed: response surface methodology (RSM) and artificial neural networks (ANN). While RSM provided interpretable regression-based surface models, the ANN models outperformed RSM in prediction accuracy, particularly in handling complex, nonlinear dependencies between the process inputs and surface outputs. To explore a practical application of FFF beyond structural parts, this research also focused on the development of electrical discharge machining (EDM) electrodes using rapid tooling (RT) techniques. Two different methods were implemented: electroplated rapid tooling (ERT) and casted rapid tooling (CRT). In both methods, polymer templates were fabricated via FFF and subsequently converted into functional copper electrodes—via electroplating in the ERT route and metal casting in the CRT route. These electrodes were then evaluated in actual EDM trials and benchmarked against a conventional solid copper (SC) electrode. The performance metrics included material removal rate (MRR), electrode wear rate (EWR), surface roughness (SR) of the machined workpiece, and tool out-of-roundness (OOR). Results revealed that the ERT electrode offered superior dimensional stability and lower wear, closely matching the performance of the SC electrode. On the other hand, the CRT electrode exhibited higher MRR but suffered from increased wear and dimensional distortion, primarily due to microstructural defects introduced during the casting process. FESEM analysis of the electrode surfaces provided microstructural evidence supporting these performance trends. The successful development of FFF-based electrodes through rapid tooling demonstrates the viability of additive manufacturing in functional tooling applications, paving the way for cost-effective and customized EDM solutions. Overall, this thesis presents a holistic framework for enhancing the performance of FFFprinted PC parts through data-driven optimization and also demonstrates the practical viability of FFF in rapid tooling for EDM applications, contributing valuable insights for the advancement of polymer-based additive manufacturing in both structural and functional domains