A. Grant, B.D. Ellis, M. Rais-Rohani
University of Maine,
Keywords: additive manufacturing, process-structure-property-performance, design of experiments
Summary:Additive manufacturing (AM) has facilitated advances in biomedical, aerospace, and fixturing applications due to reduced lead times and enhanced geometric freedom. However, these advantages come at a cost: AM imbues process-dependency into mesostructures, causing directionally-dependent material properties, reduced mean material properties, and increased uncertainty in material properties. Understanding and quantifying AM’s process-dependency and the development of process-dependent simulation techniques are paramount to continued AM advancement and industrial acceptance. This research seeks to demonstrate a systematic design of experiments approach to quantify process-dependency of structures and properties. The exploration is demonstrated via fused-filament fabrication (FFF) processing of poly-lactic acid (PLA) materials. Process-structure-property-performance (PSPP) relations cast the understanding of material characterization into a preformed framework. Process decisions give rise to structural characteristics, property expectations, and ultimately performance metrics (cf. Figure 1). In the production of an AM part, depending on part complexity, fully specifying the process involves dozens to hundreds of decisions. This research aims to explore process-structure relations, exposing which process decisions bear the most relevance on achieving a structure which gives rise to optimal achievable strength properties. A Latin hypercube design of experiments was employed to quantify the effects of extruder temperature (210°C-240°C), extrusion multiplier (100%-120%), and fan speed (0%-100%) on ultimate tensile strength (UTS) of transverse filament orientation (TFO) specimens produced using FFF (cf. Figure 2a). A total of 20 half-scale ASTM D638 Type 1 dog-bone tensile specimens  were manufactured and tested at 20 combinations of input parameter levels. Empirical data from the design of experiments were then analyzed to generate two multidimensional surrogate models via a multivariate full quadratic polynomial response surface and a multiquadric radial basis function. Both models were used to identify the optimal combinations of parameter levels to maximize UTS. Validation experiments demonstrated that the UTS of TFO specimens were increased by 42% from 40 MPa to 57 MPa. The observed increase in UTS of TFO specimens is important because the transverse strength is of similar magnitude to tensile strength in specimens with longitudinally oriented filaments (cf. Figure 2b). Results from the tensile coupons are used to inform process-dependent finite element simulations of an AM beams subject to 3-point loading and having overall dimensions of 6”x1”x1” (LxWxH). The process-dependency was estimated using a commercially-available software, employing an algorithm to account for strength under various loading conditions. In conjunction with ABAQUS, the simulations study the effects of filament orientation and thermal conditions on the flexural stiffness and strength of the beams. The presentation will show and discuss simulation results with a comparison to empirical data, thus indicating the utility and necessity of considering process-dependency in the simulation of AM part performance. This work is supportive of developing a physics-based framework that will enable designers to design functional parts for existing AM processes, and facilitate future AM innovations, such as process-dependent topology optimization, in situ quality control systems, and the design of next generation processes.