Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties

Dong Xiao *, Zahra Sharif-Khodaei, M. H. Aliabadi · 2025

Read Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties by Dong Xiao *, Zahra Sharif-Khodaei, M. H. Aliabadi (2025) 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-2501-18393.txt.

Introduction

Dong Xiao, Zahra Sharif-Khodaei, and M Aliabadi presented in 2025 a method to find where an impact hits a composite aircraft part. Heat, impact weight, and energy change in real use and can throw off older methods. Their approach uses the order of arrival times of stress waves at sensors.

A regression method with a custom kernel learns from wave travel traits and can predict both position axes at once. Several kernel results are then blended with weights that adapt to conditions. Tests on a composite plate, with both large and small falling masses, showed strong results. The method needed no fix for heat or mass changes, which suits real flight structures.

Catalog record

CREATORDong Xiao *, Zahra Sharif-Khodaei, M. H. Aliabadi
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE75 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2501-18393

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Robust impact localisation on composite aerostructures using kernel design and Bayesian fusion under environmental and operational uncertainties Dong Xiao *, Zahra Sharif-Khodaei, M. H. Aliabadi Abstract Keywords Structural health monitoring; Impact localisation; Environmental and operational uncertainties; Time difference of arrival; Gaussian process regression; Composite kernel design and fusion 0 0 footnotetext: * Corresponding author 1 1 footnotetext: Email addresses: [email protected] (D. Xiao); [email protected] (Z. Sharif-Khodaei); [email protected] (M.H. Aliabadi) 2 2 footnotetext: ORCID: D. Xiao ; Z. Sharif-Khodaei ; M.H. Aliabadi 1 Introduction Composite aircraft airframes in operational service are vulnerable to a variety of impact events, such as tool drops, hail, and bird strikes [ 1 ] . Among these, low-velocity impacts on composite aerostructures pose a significant risk of inducing Barely Visible Impact Damage (BVID) [ 2 ] , including matrix cracking [ 3 ] , delamination [ 4 ] , and fiber breakage [ 5 ] , which can severely compromise structural integrity. To effectively detect and monitor these random impacts, passive structural health monitoring (SHM) techniques can be integrated into the airframe [ 6 ] , enabling early identification of the location, presence, and severity of impact-induced damage. A comprehensive passive SHM system comprises two key components: passive sensors (hardware) permanently affixed to or embedded within the composite structures, and impact identification algorithms (software) that provide real-time localisation and estimation of impact force and damage severity. Among these processes, impact localisation is fundamental, as it provides the essential spatial information required for subsequent damage assessment and mitigation. Impact localisation is an inverse problem aimed at determining the location of structural excitation (inputs) from structural outputs (passive sensor signals). Significant research efforts have been devoted to developing real-time impact localisation methodologies for composite aerostructures, considering their anisotropic and complex material properties. These methodologies are broadly classified into two categories: model-based methods and data-driven methods. Model-based methods rely on physical and mathematical representations of impact dynamics, including finite element analysis (FEA) [ 7 , 8 , 9 , 10 ] , structural modal analysis [ 11 , 12 , 13 ] , and structural transfer functions [ 14 , 15 , 16 , 17 , 18 , 19 ] . These methods excel in controlled conditions but often require precise knowledge of material properties [ 20 ] and boundary conditions [ 21 ] , which can be challenging to obtain in real-world applications. Ensuring their accuracy involves iterative model updating [ 22 ] , validation, and verification [ 23 ] . In contrast, data-driven methods rely on surrogate models constructed from reference databases to map structural outputs (e.g., sensor signals) to impact inputs (e.g., location and force). Traditional approaches, such as multi-layer perceptrons (MLPs) [ 24 , 25 ] and Gaussian process regression (GPR) [ 26 , 27 ] , depend on features like time differences of arrival (TDOA) [ 28 , 29 ] , rooted in wave propagation principles [ 30 , 31 , 32 , 33 , 34 , 35 , 36 ] . Modern deep learning techniques, including convolutional neural networks (CNNs) [ 37 , 38 ] , temporal convolutional networks (TCNs) [ 39 ] , recurrent neural networks (RNNs) [ 40 , 41 ] , and graph neural networks (GNNs) [ 42 , 43 ] , directly process raw time-series data, capturing complex patterns and nonlinear relationships. However, deep learning methods often demand extensive, high-quality datasets that are challenging to obtain for real-world SHM applications. A critical challenge in both model-based and data-driven approaches is their limited ability to generalise from controlled laboratory conditions to real-world operational environments. Real-world variability, including temperature fluctuations, impact mass, energy, and environmental noise, can significantly degrade the performance of impact localisation models. For instance, even minor temperature variations can alter wave propagation velocities, introducing errors in time difference of arrival (TDOA)-based localisation. Furthermore, substantial changes in impact mass dramatically modify the impact responses, resulting in entirely different spectral distributions of sensor signals. While Seno [ 44 , 45 ] experimentally demonstrated that small variations in TDOA occur due to changes in impact mass, height, and angle with the same impactor, these studies were limited to minor perturbations. Such findings are insufficient to address the challenges posed by significant variations in impact mass and environmental temperature, which require more robust and adaptable localisation methodologies. To address these challenges, recent studies have explored techniques such as data augmentation [ 46 ] , transfer learning [ 47 ] , and hybrid physics-data learning [ 48 ] . While promising, these methods face limitations in replicating the full spectrum of real-world uncertainties (data augmentation and transfer learning) or integrating physics-based and data-driven insights effectively (hybrid physics-data learning). This study introduces a novel approach that leverages the inherent invariance in the order of TDOA under environmental and operational uncertainties. For example, temperature variations primarily scale wave propagation velocities proportionally, thereby preserving the relative order of TDOA across sensors. Additionally, substantial changes in impact mass alter the frequency spectrum of impact responses, enabling TDOA extraction at different dominant frequencies, which scale proportionally due to bending wave dispersion. Building on these insights, a composite (COMP) kernel is developed for GPR, combining the strengths of the radial basis function (RBF) and cosine similarity (COS) kernels. The RBF kernel effectively captures nonlinear relationships, while the COS kernel models linear and angular dependencies in TDOA, ensuring enhanced robustness to environmental and operational variability. To further improve adaptability, Bayesian model averaging (BMA) is employed to dynamically fuse these kernels, assigning context-sensitive weights that enhance overall model reliability and accuracy. Extensive experimental validation was conducted on a fiber-reinforced composite panel using a drop tower machine and a guided drop mass across a wide range of impact conditions. These conditions encompassed significant environmental and operational uncertainties, including variations in impact mass, energy, angle, and temperature. The proposed method demonstrated exceptional robustness and generalisability under these diverse scenarios, highlighting its potential for real-world structural health monitoring applications. The key contributions of this study include: • Robust kernel design: Development of a novel composite kernel for GPR, integrating RBF and COS kernels. This design leverages wave propagation dynamics to enhance robustness against environmental and operational uncertainties. • Bayesian kernel fusion: Implementation of Bayesian model averaging (BMA) to dynamically combine kernel predictions, enabling adaptive kernel weighting for improved accuracy and reliability across diverse impact conditions. • Comprehensive experimental validation: Rigorous validation on a fiber-reinforced composite panel under a wide range of conditions, including variations in impact mass, energy, angle, and temperature, showcasing the generalisability and practical applicability of the proposed method. • Sample Standardisation: Demonstration of the superiority of sample standardisation (SS) over feature standardisation (FS) for preprocessing TDOA inputs. SS preserves order invariance and enhances compatibility with GPR models under varying uncertainties. • Independence from Boundary Conditions: Experimental evidence demonstrating the effectiveness of the proposed TDOA-based impact localisation method irrespective of structural boundary conditions. The remainder of this paper is organised as follows: Section 2 provides an overview of passive SHM sensing for impact monitoring on aircraft structures. Section 3 discusses the wave propagation relationship between impact location, wave velocity, and TDOA, along with the influence of environmental and operational uncertainties, which are critical to localisation accuracy. Building on this foundation, Building on this

Read the complete text