Additive Manufacturing Process Optimization and Defect Characterization

Additive manufacturing (AM) encompasses a suite of layer-by-layer fabrication techniques that enable the rapid production of complex geometries from digital models. Central to the industrial adoption of AM is the optimisation of process parameters—such as laser power, scan speed, layer thickness and powder characteristics—to achieve target densities, microstructures and mechanical properties. Variations in thermal gradients, material feed and scan strategies give rise to characteristic defects, including lack-of-fusion pores, keyhole pores and gas-entrapped inclusions. Defect characterisation employs non-destructive evaluation tools—principally X-ray computed tomography (XCT)—to visualise internal porosity and macro-scale distortions, while in situ sensors and image-based simulations offer real-time feedback for closed-loop control. Recent advances in machine learning and data-driven reconstruction have further accelerated defect detection and quantitative analysis, permitting high-throughput evaluation of component batches. Optimization efforts now focus on integrating multi-scale modelling, advanced imaging and artificial intelligence to minimise critical defects and ensure reproducibility across materials and geometries. This convergence of process control and defect analytics underpins the global drive towards qualified, defect-tolerant additive workflows for aerospace, biomedical and high-value manufacturing sectors.

Research from Nature Portfolio

Recent studies have introduced a feature-based methodology for volumetric defect classification in powder-bed fused Ti-6Al-4V. High-resolution XCT data were processed to extract nine morphological parameters—such as maximum dimension, aspect ratio and sparseness—for three common defect types: lack-of-fusion pores, gas-entrapped pores and keyholes. Statistical analysis revealed significant parameter overlap, prompting the development of a multi-parameter classification scheme. When implemented in decision-tree and neural-network frameworks, the approach achieved over 98 % and 99 % accuracy respectively, enabling robust, automated discrimination of defect populations and paving the way for real-time quality assessment in metal AM.

Research from all publishers

A deep learning-guided approach has been proposed to accelerate XCT acquisition and reconstruction for metal AM parts. By integrating computer-aided design models with physics-based simulations of X-ray interactions, a convolutional network was trained to correct beam-hardening and other artefacts, reducing scan times while preserving dimensional and density resolution. High-throughput characterisation of over one hundred AlCe alloy components demonstrated marked improvements in speed and image fidelity, facilitating rapid assessment of process-parameter effects on porosity distribution.

An explainable artificial intelligence (XAI) framework has been developed for automatic detection and characterisation of defects in Ti-6Al-4V specimens. Support vector machines classify CT pixel data into pores and inclusions with an area-under-curve exceeding 0.94, while density-based clustering groups defect pixels for morphological analysis via convex-hull metrics. This methodology matches expert inspection and delivers intuitive visualisations that underpin trust in automated NDT workflows.

A material-agnostic strategy for pore-type classification exploits geometric features that remain invariant across powder-bed fusion scenarios. By analysing distributions of shape descriptors—such as sphericity, surface curvature and elongation—across multiple alloys and laser settings, a supervised classifier was trained to distinguish keyhole, lack-of-fusion and gas pores. The model achieved up to 93 % accuracy in single-material tests and 90 % in cross-material validation, offering a generalised tool for pore identification in diverse AM processes.

Additive Manufacturing Process Optimization and Defect Characterization publication trend

The graph below shows the total number of articles in additive manufacturing process optimization and defect characterization across all publications each year (not limited to Nature Index journals).

Technical terms

Additive Manufacturing (AM): A collection of layer-wise fabrication methods for building parts directly from digital models.

Powder-Bed Fusion: An AM process that selectively fuses powder layers using a heat source such as a laser or electron beam.

X-ray Computed Tomography (XCT): A non-destructive imaging technique that reconstructs three-dimensional internal structures from X-ray projections.

Porosity: The presence of voids or pores within a material, often detrimental to mechanical performance.

Keyhole Pore: A defect formed by unstable melt-pool dynamics, characterised by elongated, irregular voids.

Explainable Artificial Intelligence (XAI): AI methods that provide interpretable and transparent decision logic alongside predictive performance.

Deep Learning-Based Reconstruction: The use of neural networks to correct imaging artefacts and accelerate image reconstruction workflows.

References

1. Enabling rapid X-ray CT characterisation for additive manufacturing using CAD models and deep learning-based reconstruction. npj Computational Materials (2023).
2. eXplainable artificial intelligence for automatic defect detection in additively manufactured parts using CT scan analysis. Journal of Intelligent Manufacturing (2023).
3. Towards material and process agnostic features for the classification of pore types in metal additive manufacturing. Materials & Design (2023).
4. Feature-based volumetric defect classification in metal additive manufacturing. Nature Communications (2022).