Full-scale modal testing of a Hawk T1A aircraft for benchmarking vibration-based methods

TL;DR

This paper presents a comprehensive, openly available benchmark dataset from full-scale modal testing of a Hawk T1A aircraft, aimed at advancing structural dynamics research and validation of identification and prediction methods.

Contribution

It provides a detailed dataset from a full-scale aircraft test, including experimental design, data collection, baseline results, and discussion of challenges for structural dynamics benchmarking.

Findings

Dataset includes frequency and time domain data

Baseline results for identification and prediction tasks

Discussion of dataset-specific challenges

Abstract

Research developments for structural dynamics in the fields of design, system identification and structural health monitoring (SHM) have dramatically expanded the bounds of what can be learned from measured vibration data. However, significant challenges remain in the tasks of identification, prediction and evaluation of full-scale structures. A significant aid in the roadmap to the application of cutting-edge methods to the demands of in-service engineering structures, is the development of comprehensive benchmark datasets. With the aim of developing a useful and worthwhile benchmark dataset for structural dynamics, an extensive testing campaign is presented here. This recent campaign was performed on a decommissioned BAE system Hawk T1A aircraft at the Laboratory for Verification and Validation (LVV) in Sheffield. The aim of this paper is to present the dataset, providing details on the structure, experimental design, and data acquired. The collected data is made freely and openly available with the intention that it serve as a benchmark dataset for challenges in full-scale structural dynamics. Here, the details pertaining to two test phases (frequency and time domain) are presented. So as to ensure that the presented dataset is able to function as a benchmark, some baseline-level results are additionally presented for the tasks of identification and prediction, using standard approaches. It is envisaged that advanced methodologies will demonstrate superiority by favourable comparison with the results presented here. Finally, some dataset-specific challenges are described, with a view to form a hierarchy of tasks and frame discussion over their relative difficulty.