Dynamic Deformation Reconstruction of Variable Section WING with Fiber Bragg Grating Sensors

Zhen Fu, Yong Zhao, Hong Bao, Feifei Zhao2019 · DOI

TL;DR

This paper introduces a novel dynamic deformation reconstruction algorithm for variable-section wings using Fiber Bragg Grating (FBG) sensors. By combining an inverse finite element method with a self-structuring fuzzy neural network for strain correction, the approach achieves high-precision real-time shape sensing, crucial for advanced aircraft control.

What Problem Does This Paper Address?

Accurately monitoring the real-time deformation of aircraft wings is vital for flight safety and intelligent control systems. While Fiber Bragg Grating (FBG) sensors offer advantages for shape sensing, reconstructing wing deformation from strain data is challenging due to factors like sensor installation errors and dynamic un-modeled errors (differences between static and dynamic models). Existing methods often struggle with variable-section structures or dynamic loads. This paper aims to address these limitations by proposing a robust algorithm that can accurately reconstruct dynamic deformations of variable-section wings, even in the presence of these errors.

Methods

The core of the proposed method lies in a two-pronged approach:

  • Inverse Finite Element Method (iFEM) for Variable-Section Beams: The authors first establish a deformation reconstruction model for a variable-section beam element based on Timoshenko beam theory and the iFEM framework. This model relates measured strain to structural deformation.
  • Self-Structuring Linear Support Vector Regression Fuzzy Network (SSILSVRFN) for Strain Correction: To mitigate errors arising from sensor placement and dynamic un-modeled effects, a sophisticated fuzzy network is employed. This network is trained using support vector regression (SVR) to learn the complex relationship between measured and actual strain. The SSILSVRFN is designed with a self-structuring capability, allowing it to adapt its rules and structure based on the input data, thereby improving generalization and preventing overfitting. The network learning involves both structure learning (determining fuzzy rules) and parameter learning (optimizing antecedent and consequent parameters).

Datasets

  • Simulated Data: A finite element (FE) model of a variable-section wing was created using ABAQUS software. This model was used to generate strain data under different loading conditions for evaluating the iFEM reconstruction model.
  • Experimental Data: A physical variable-section wing model was tested under dynamic loading conditions. Strain data was collected using FBG sensors, and actual deformation was captured using a 3D optical measurement instrument (NDI Optrotrak Certus). This experimental setup was used to validate the performance of the complete proposed algorithm.

Key Insights and Ideas

  • Strain Correction is Crucial for Dynamic Loads: The study highlights that dynamic un-modeled errors and sensor placement inaccuracies significantly impact deformation reconstruction accuracy, especially under dynamic loading. The SSILSVRFN effectively addresses these issues by learning and correcting the measured strain data.
  • Adaptive Fuzzy Networks for Complex Relationships: The SSILSVRFN's ability to adapt its structure and rules based on data is key to its success in modeling the non-linear and complex relationship between measured and actual strain, which is difficult to capture with traditional mathematical models.
  • iFEM for Variable Sections: Extending the iFEM framework to handle variable-section beam elements is essential for accurately modeling complex wing geometries.
  • Integrated Approach for Robustness: The combination of a physics-based model (iFEM) with a data-driven learning approach (SSILSVRFN) provides a powerful and robust solution for dynamic deformation reconstruction.

Key Results

  • Simulation Accuracy: In FE simulations, the proposed iFEM reconstruction model achieved high accuracy, with percent errors for the main deformation direction (wz) remaining below 6.0%.
  • Experimental Accuracy (Uncorrected): Without strain correction, experimental studies showed percent errors in deformation reconstruction below 13%.
  • Experimental Accuracy (Corrected): After applying the SSILSVRFN strain correction, the experimental results demonstrated a significant improvement, with percent errors for the main deformation direction remaining below 6.7%.
  • RMS Error Reduction: The Root Mean Square (RMS) error for deformation reconstruction was reduced by at least 66.8% after applying the SSILSVRFN correction. For example, the RMS error for deformation along the y-axis decreased from 12.01 mm to 3.97 mm at one time point.

Limitations and Future Work

  • Generalizability: The authors note that if the measurement objects are different, the reconstruction model and the correction network need to be re-established. This implies that the current model is specific to the tested wing geometry and loading conditions.
  • Dynamic Load Simulation: The dynamic loading in the experiment was caused by the sudden removal of a load. The authors suggest future work should involve performing dynamic experiments on a shaking platform to simulate dynamic deformations caused by different airflow excitations, which would provide a more comprehensive assessment of the algorithm's effectiveness.

Summary updated

BibTeX
@misc{fu2019dynamic,
  author = {Zhen Fu and Yong Zhao and Hong Bao and Feifei Zhao},
  title = {{Dynamic Deformation Reconstruction of Variable Section WING with Fiber Bragg Grating Sensors}},
  year = {2019},
  doi = {10.3390/s19153350},
}

Loading related content...

Figure