Issue 77

T. Hachimi et alii, Fracture and Structural Integrity, 77 (2026) 173-206; DOI: 10.3221/IGF-ESIS.77.11

that natural surface textures in PBF processes can be of sufficient contrast to allow meaningful correlation without the need for artificial patterning. Diani et al. [25] directly address this by substituting ink-stamped circular dots for the speckle in a reproducible manner, allowing for better control of speckle morphology, whilst Liu et al. [60] further develop microscale speckle that is resistant to heat for high-temperature AM environments. Wang and Lei [108] utilize deep learning to super resolve DIC data, allowing for enhanced correlations from lower-res images. With this awareness, specific wood-fiber filaments exist that can form natural speckles during printing processes, as demonstrated by Holzmond and Li [46]ndeed, this trend is towards the more “pattern-agnostic” types of algorithms, which are generally surfacing now. Szalai et al. [100] also show that surface cleaning and priming parameters are just as important as the patterning step itself for effective correlation of complicated structural materials. For conventional macro-scale methods, Dong and Pan [26] show in their proof-of-principle that conventional scratching using sandpaper or dry toner can be used, and by looking at the relative distributions of the greyscale images between the reference and distorted counterparts, the entire displacement field can be reconstructed (Figure 3).

Figure 3: Schematic representation of DIC subset tracking and deformation gradient mapping across a stochastic speckle pattern. Metrological constraints and spatial resolution In AM polymer characterisation, a central challenge is the critical trade-off of subset sizes, measurement noise, and resolution [6,91]. The DIC spatial resolution is roughly proportional to subset size, and for first-order shape functions, the displacement resolution is approximately proportional to ~(2M+2) for a subset of (2M+1)×(2M+1) pixels [91]. In polymers like PLA and ABS, Acciaioli et al. [2] and Perez et al. [85] as localised strain occurs at the ‘neck’ between deposited beads, inappropriately large subsets smooth over critically needed data and severely underestimate the peak strains crucial for proper fracture modelling. As a result of such considerations, we may find in literature that the technical depth is absent in studies that do not report the Metrological Efficiency Indicator (MEI) or, at the very least, perform baseline noise studies to obtain the limiting strain detection capabilities, the importance of which is evident for reliably measuring small elastic strains in stiff, high-performance polymers [91]. A useful summary of such algorithm parameters as they influence the AM characterisation is given in Table 1.

DIC Algorithm Parameter

Impact on AM Characterization

Technical Implication

Larger subsets smooth over filament-level strain concentrations [91] Smaller steps increase the resolution of localized shear bands [2] Higher-order functions are needed for complex lattice strut buckling [2,6] ZNSSD is required for in-situ monitoring with flickering heat sources [2,83]

Subset Size

Signal-to-noise ratio

Step Size

Strain map density

Shape Function Order

Accuracy of warping

Correlation Criterion

Lighting invariance

Table 1: DIC algorithm parameters and their direct impact on AM polymer characterization accuracy.

178