Identifiability Limits in Ultrasonic Microstructure Characterisation: A Canonical and Stochastic Framework
Wei Yi Yeoh · 2026
Read Identifiability Limits in Ultrasonic Microstructure Characterisation: A Canonical and Stochastic Framework by Wei Yi Yeoh (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2605-13057.txt.
Introduction
Yeoh asks how much ultrasound can truly reveal about material microstructure, a question of identifiability rather than inversion success alone. For a standard pulse-echo model, a closed-form study shows limits set by linked parameters, tight dimension caps and interface saturation. For random-field stand-ins the map from inputs to outputs stays full rank, yet its sensitivity is lopsided. Real variability cuts what can be pinned down further, though pooling observables sharpens the picture via complementary sensitivities.
Catalog record
| CREATOR | Wei Yi Yeoh |
|---|---|
| YEAR | 2026 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 65922 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2605.13057
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Identifiability Limits in Ultrasonic Microstructure Characterisation: A Canonical and Stochastic Framework Wei Yi Yeoh Abstract Ultrasound for microstructure characterisation is increasingly studied and is often assessed through inversion performance. However, the framework is fundamentally constrained by the information content available in the measured response. Hence, this work examines identifiability directly by analysing the geometry of the forward operator in both a canonical pulse-echo model and a stochastic surrogate microstructure. For the canonical model, a closed-form sensitivity analysis reveals information limits arising from parameter coupling, dimensional restriction, and interface-driven saturation. For the surrogate microstructures represented by Gaussian random fields, the forward map from correlation length D D and texture-coherence parameter T T to the attenuation and velocity observables remains structurally full rank. However, the sensitivity geometry is strongly anisotropic, with uneven parameter influence across the observable space. When intrinsic microstructural variability is incorporated, practical identifiability is further reduced. A variance-weighted Fisher framework shows that recoverability is governed by the balance between sensitivity magnitude and stochastic variability, rather than by structural rank alone. Inversion results confirm this behaviour: single observables produce elongated and weakly constrained objective landscapes, whereas combined observables improve conditioning through complementary sensitivities. These results show that, within the feature-level framework considered here, identifiability limits are governed primarily by forward-map structure and intrinsic variability, with direct implications for observable selection and measurement design. [type=editor] keywords Ultrasound ,Identifiability ,Wave Scattering ,Sensitivity Analysis ,Polycrystalline Materials ,Microstructure Characterisation † † corresponding: Corresponding author 1 Introduction Ultrasound is increasingly used for non-destructive evaluation (NDE) and the characterisation of material microstructure. The elastodynamic principle couples the effects of wave propagation and scattering to variations in material properties, enabling the inference of microstructure information based on ultrasound response [ 51 ] . This relationship provides the basis for ultrasonic microstructure characterisation methods that span from analytical scattering models to numerical simulations and experimental inversion approaches. The theoretical framework for ultrasonic scattering in polycrystalline materials was established through the unified scattering theory that relates attenuation and velocity dispersion to elastic heterogeneity [ 58 , 70 ] . The general formulation was later adapted to evaluate backscatter grain noise [ 45 , 46 ] . Since then, scattering models have been extended to include more realistic microstructural descriptors, such as second-order scattering effects [ 61 ] , elongated grain morphologies [ 31 , 38 ] , grain size distributions [ 49 , 55 ] , anisotropic texture [ 67 ] , multi-phase microstructures [ 44 ] , and wave modes [ 33 ] . These studies show that ultrasonic observables are sensitive to microstructural features, with some inherent coupling effects between physical mechanisms. In parallel, numerical modelling has become an important framework to validate the scattering theories [ 62 , 9 ] and to explore regimes where analytical assumptions become restrictive. Finite element (FE) simulations have been used to study wave propagation and scattering in two- and three-dimensional polycrystalline media [ 6 , 68 ] that contain explicit grain morphologies generated through tessellation methods [ 48 ] . Apart from the addition of complex features such as elongated grains [ 4 , 19 ] , highly anisotropic and scattering properties [ 18 , 11 ] , textured heterogeneities [ 30 , 72 ] , and multi-phase material [ 43 ] , recent work also used generated synthetic microstructures derived from experimental microstructure data to improve the representation of simulated media [ 12 , 73 ] . This modelling framework has made it possible to examine the influence of microstructure descriptors on ultrasonic responses that otherwise would be difficult to isolate experimentally. Experimental approaches have been progressively adopted to validate the analytical and numerical models and to infer microstructural descriptors directly from ultrasonic measurements. Attenuation-based approaches are used to estimate grain sizes in metallic alloys [ 41 , 53 ] . Backscatter-based approaches are used to characterise spatially varying microstructures [ 35 ] and anisotropic heterogeneity [ 32 , 65 ] . Velocity-based approaches are used to evaluate elastic stiffness [ 25 ] , crystal orientations [ 10 , 59 ] , and bulk texture [ 26 ] . Recent studies have increasingly used phased arrays to improve backscatter-based microstructure characterisation, where beamforming enables spatially resolved and directionally selective interrogation of scattering signatures [ 40 , 69 , 39 ] . Beyond conventional measurements, alternative ultrasonic modalities have also been explored for microstructure characterisation, including resonant ultrasound spectroscopy for inferring effective elastic properties [ 27 ] , grazing-incidence ultrasonic microscopy for grain-scale surface interrogation [ 22 ] , and coda-wave method that exploits multiply scattered late-time signals to estimate grain-sizes [ 15 ] . In recent years, machine-learning approaches have become increasingly popular in ultrasound. Particularly for materials characterisation, data-driven networks are used to identify grain-size classes from backscattered data [ 66 ] and to infer grain-size distributions from non-linear wave responses [ 36 ] . A similar physics-guided simulation approach enabled the acceleration of anisotropic weld properties reconstruction from array data [ 57 ] and crystallographic orientation mapping from surface wave measurements [ 47 ] . Physics-informed networks have also been explored for inferring spatially varying elastic properties from wavefield data and demonstrated potential for solving ill-posed inverse problems [ 56 ] . While these studies demonstrate the improved analysis and acceleration with the use of machine-learning networks, they do not eliminate the need to understand whether the measured wavefields contain sufficient independent information about the target descriptors. These developments demonstrate the increasing maturity of ultrasonic microstructure characterisation. However, they also highlight a central challenge: in practical engineering materials, multiple descriptors often influence the same measured wavefield in coupled and anisotropic ways. Recent studies have begun to address this complexity through multi-parameter analysis [ 50 , 14 ] . Nevertheless, it is often unclear whether the ambiguities in the inferred properties arise from limitations of the inversion method or from intrinsic non-uniqueness in the forward mapping. This motivates an identifiability-based analysis of ultrasonic microstructure characterisation. Rather than asking only whether an inversion or learning method can reproduce a target descriptor, the central question is whether the chosen ultrasonic observables contain sufficient independent and stable information to recover that descriptor in the first place. In this work, a unified identifiability study is presented to quantify the intrinsic information content of ultrasonic measurements through the analysis of a feature-level forward map [ 7 , 8 , 13 ] . A minimal canonical model is first used to isolate the fundamental mechanisms governing sensitivity and parameter coupling, before extending the same analysis to stochastic surrogate microstructures that capture distributed scattering and variability. Across both settings, identifiability is characterised through singular-value structure and a variance-weighted Fisher framework. The results show that the main challenge to identifiability is not only dimensional restriction, but also imbalance in sensitivity across parameter directions. In the surrogate setting, the forward map remains structurally full rank, yet certain directions are intrinsically weak and become further suppressed when stochastic variability is taken into account. This leads to a loss of practical identifiability, showing that the limits of ultrasonic inversion are governed by the geometric structure of the forward map and the information content available in the measurement. The contributions of this paper are threefold. First, a geometric baseline for