This work explores the synergy between metal additive manufacturing and topology optimization to enable lightweight structures with superior mechanical properties. The study first proposes a novel multi-scale topology optimization methodology to generate lattice-embedded structures, and applies it to the design of a bridge-like part under three-point bending. Four designs are generated, spanning macroscale volume fractions from 15% to 30% to explore mass–stiffness trade-offs. Each design is manufactured four times in a high strength aluminum alloy using laser powder bed fusion. Although pragmatic industrial examples exist, to the authors’ knowledge, this is the first reported study in which topology-optimized, lattice-embedded metal structures were successfully fabricated. Experimental three-point bending tests assess printing quality, the force–displacement response (linear and nonlinear) and failure behavior. The results show excellent print fidelity, low variability, and close agreement between measured and numerical volume fractions. While linear stiffness does not match numerical predictions, the experiments reveal highly tunable crack propagation. In particular, the sequential failure of unit cells leads to stable energy absorption and placement of solid regions is shown to cause direction-changing crack paths. The findings highlight the potential of jointly optimizing lattice geometry and solid–lattice distribution to tailor mechanical response and failure modes in additively manufactured structures.
Crack formation of CM247LC makes fabrication of this alloy by powder bed fusion-laser beam a major challenge. Rather than focusing on prevention strategies, this work investigates the possibility of monitoring crack formation using in-situ acoustic measurements. During printing, impulsive acoustic events are observed in structure-borne acoustic emission recordings, and a signal processing pipeline is proposed in order to automatically detect these events over the full print. No clear correlation was found between the number of these acoustic events and the observed crack density. However, the acoustic impulses are hypothesized to be caused by post-solidification cracks, where either the hot crack or the sharp edges of lack of fusion porosities act as initiation sites for small scale cracks. This is further validated using operando synchrotron imaging, showing that the formation of a hot crack does not seem to generate acoustic impulses, but the morphology change of the crack after solidification does. Finally, these acoustic events were not detectable with an optical microphone.
Powder recoating is the first step of every powder bed fusion (PBF) layer, and material protrusions rising above the nominal powder surface can obstruct the recoater and compromise an entire build. Here we demonstrate a novel method for protrusion detection and localization from the structure-borne acoustic signal of the recoater blade itself. Two piezoelectric elements are bonded to the ends of a thin comb-shaped stainless-steel blade, so that the blade acts as the diaphragm and the elements transduce the dynamic strain at its surface. Spectral analysis shows that the recorded signal is dominated by flexural waves excited when a tooth is deformed, with a dominant frequency corresponding to the second cantilever mode of a single tooth. Times of arrival were picked per channel with a median-referenced energy-ratio detector, and the resulting time difference of arrival scales linearly with the difference in source-to-sensor distance across the blade width. Using this calibration, artificial protrusions raised in 0.1 mm increments were detected and localized, in agreement with optical images of the powder bed.
The mechanical performance of additively manufactured glass remains sparsely explored, and few design strategies for strengthening printed glass have been demonstrated. This study asks if complex geometries enabled with an extrusion-based additive manufacturing process can improve the mechanical performance of a soda-lime glass object. Hollow single-wall cylinders were printed using a vase-mode approach, applying internal reinforcing patterns inspired by the Euplectella aspergillum (Venus’ flower basket) glass sponge and a hexagonal honeycomb. Under lateral compression both patterns raised mean breaking load over an unreinforced control specimen, by 37.1% for honeycomb and 27.4% for Venus’ flower basket. Normalised by mass, the improvements were 17.6% and 5.3% respectively. Finite element analysis of simplified geometries led to a postulate that patterns concentrate stress into small regions, possibly reducing the effective flaw sampling area and volume. Weibull size scaling from feedstock rods overpredicted measured strength, indicating that deposition introduced failure mechanisms beyond the flaw population of the feedstock and that behaviour-specific empirical testing is required for an accurate understanding of performance.
Accurate heat source calibration is essential for reliable Laser Powder Bed Fusion (LPBF) Additive Manufacturing (AM) process simulation, yet conventional case-by-case calibration remains computationally expensive and experimentally intensive. This study presents a machine learning (ML)-assisted framework for predicting the calibrated heat source parameters required for melt pool modeling in LPBF across Nickel-based superalloys. A comprehensive dataset was established by integrating literature-derived data with experimentally calibrated melt pool measurements for Inconel 718, Inconel 625, and Hastelloy X over the investigated processing window. The proposed ML models were trained to map process and material descriptors to the regression coefficients that define the heat source penetration depth (H) and effective laser absorptivity (A) within a conical-Gaussian heat source model used to represent the conduction regime of the melt pool. Among the evaluated algorithms, the neural network (NN) demonstrated the best predictive capability, achieving coefficients of determination (R2) greater than 0.90 for the H and A values derived from the predicted coefficients while reducing calibration time from approximately 35 min in conventional simulation workflows to a few seconds. Incorporating material chemical composition as an input feature further improved prediction accuracy for the three investigated Nickel-based superalloys. Within the investigated alloys and processing window, the developed framework offers an efficient and physically consistent alternative to repetitive heat source recalibration and establishes a practical route for accelerated process optimization and efficient multiscale thermal modeling in LPBF.
