Improved deformation reconstruction of composite material structures based on optical fiber sensing Authors: Jun Huang, Yuwei Sun, Yu Peng, Dongsheng Zhang, Ruiya Li, Sheng Yin, Zhenfeng Peng, Quanyong Huang, Quan Liu Published: 2025 ABSTRACT Due to high strength and lightweight, composite material structures have been widely used in various industries such as aerospace, automotive, civil engineering, and healthcare. Real-time deformation monitoring based on optical fiber sensing technology holds immense potential in composite materials’ structural health monitoring (SHM). In this study, structural deformation reconstruction algorithms have been investigated by improving the slope recursion algorithm and the Ko displacement theory. The slope recursion algorithm is enhanced through angle averaging. Cross-validation is employed to increase strain-fitting accuracy for the Ko displacement theory. Finite element simulation analysis and deformation reconstruction experiments of a carbon fiber composite plate were conducted to validate the deformation reconstruction algorithms. Fiber Bragg grating (FBG) sensors were applied to measure quasi-distributed strain as inputs for the algorithms. The results show that the overall reconstruction accuracy increased by 1.91% with the improved slope method and by 0.92% with the optimized Ko displacement theory. Deformation reconstruction experiments of a composite material reinforced structure were carried out, and 70 FBG sensors were installed on its surface for strain field measurement. The experimental results indicate the feasibility of the proposed deformation reconstruction methods for complex structures. Introduction Carbon fiber-reinforced composites have become indispensable structural materials in aerospace, automotive, civil engineering, and healthcare industries due to their exceptional specific stiffness, strength, corrosion resistance, and vibration damping characteristics1. Their widespread adoption is exemplified by the Boeing 787 and Airbus A350, where composites constitute approximately 50% and 52% of the airframe weight, respectively, including critical load-bearing components such as wings and fuselages2. However, the inherent anisotropy and structural complexity of composite materials render them susceptible to internal damage mechanisms—including delamination, fiber breakage, and matrix cracking—that often remain undetectable through visual inspection yet can precipitate catastrophic failures. Consequently, deformation monitoring has emerged as a crucial component of structural health monitoring (SHM), enabling damage assessment, residual life prediction, and operational safety assurance3,4. Traditional deformation monitoring techniques, primarily strain gauges and digital image correlation (DIC), exhibit significant limitations for composite applications. Although strain gauges provide localized strain measurements and have been implemented in minimally invasive forms for aerospace surfaces, they suffer from electromagnetic interference, temperature sensitivity, and complex wiring requirements for large-scale deployment5,6. Similarly, DIC offers full-field measurement capabilities and has been employed to characterize nanocomposite deformation at multiple scales, but its accuracy diminishes under poor lighting conditions or on surfaces with inadequate texture7,8. Fiber optic sensing technology has consequently gained prominence as a transformative monitoring solution, combining intrinsic advantages such as small size, immunity to electromagnetic interference, and compatibility with embedded integration9. Notably, Fiber Bragg Grating (FBG) sensors have been successfully surface-mounted on wing skins for impact damage detection10 and embedded within carbon composites for in-situ SHM11. The effectiveness of FBG-based monitoring, however, fundamentally depends on the deployment of robust deformation reconstruction algorithms to convert discrete strain measurements into displacement fields. Four primary methodologies have been developed: The inverse Finite Element Method (iFEM) offers model-independent reconstruction using strain data without requiring loading conditions, making it suitable for complex structures like wind turbine towers. Nevertheless, its accuracy is constrained by mesh design and suffers from singularity issues in irregular geometries, as addressed through enhanced formulations like the inverse local radial point interpolation method12,13. Modal superposition methods utilize strain mode shapes for deformation mapping and have been improved via transfer learning to enhance wing reconstruction accuracy14,15. Their performance, however, relies heavily on accurate finite element models and is limited under nonlinear deformations. The Ko displacement theory provides a computationally efficient solution for beam-like structures through strain integration and has been applied in composite cantilever beams using distributed sensing16,17. Its applicability remains restricted to unidirectional bending and requires dense sensor placement for reasonable accuracy. Curvature-based methods reconstruct deformed shapes by converting strain to curvature and performing geometric integration. Although advanced implementations have incorporated directional correction and Runge-Kutta solvers for three-dimensional wing deformation18,19, these methods are generally limited to unidirectional reconstruction and require significant sensor counts. Each algorithm exhibits particular limitations: iFEM struggles with mesh dependency and singularities; modal methods require precise numerical models and are unsuitable for nonlinear responses; Ko theory is restricted to simple bending; and curvature methods lack generality for complex deformations. A critical common shortcoming is their susceptibility to error accumulation at structural boundaries and limited accuracy under complex support conditions. Consequently, numerous scholars have investigated methods to enhance the accuracy and stability of these algorithms, proposing various improvement strategies. For instance, Zhao et al.20 integrated isogeometric analysis into the conventional iFEM algorithm, reducing the deformation reconstruction RMSE of a variable-cross-section panel-beam structure under uniform load from 2.78 mm to 1.34 mm. Esposito et al.21 introduced a hybrid shell-beam inverse finite element method for the shape sensing of stiffened thin-walled structures, which achieved an accuracy improvement of 1–1.2% compared to a pure shell-element model. Chen et al.22 developed a modified modal superposition method for an aircraft tooling positioner, reducing the absolute error in deformation reconstruction from 0.133 mm to 0.059 mm compared to the traditional modal superposition approach. Li et al.23 proposed a curvature-optimized deformation reconstruction method, which decreased the relative error by 3.3%, with the maximum absolute error reducing from 3.743 mm to 2.627 mm. For improving the accuracy of deformation reconstruction, this paper makes the following contributions: A novel data fitting method based on endpoint error minimization and cross-validation is proposed to improve strain measurement accuracy at critical edge regions.A recursive weighted reconstruction algorithm is developed to suppress error propagation in fixed-end structures.The proposed methods are validated through comprehensive experiments on composite specimens with varying complexity, demonstrating significant improvement in reconstruction accuracy compared to conventional approaches. A novel data fitting method based on endpoint error minimization and cross-validation is proposed to improve strain measurement accuracy at critical edge regions. A novel data fitting method based on endpoint error minimization and cross-validation is proposed to improve strain measurement accuracy at critical edge regions. A recursive weighted reconstruction algorithm is developed to suppress error propagation in fixed-end structures. A recursive weighted reconstruction algorithm is developed to suppress error propagation in fixed-end structures. The proposed methods are validated through comprehensive experiments on composite specimens with varying complexity, demonstrating significant improvement in reconstruction accuracy compared to conventional approaches. The proposed methods are validated through comprehensive experiments on composite specimens with varying complexity, demonstrating significant improvement in reconstruction accuracy compared to conventional approaches. The remainder of this paper is organized as follows: “Mathematical modeling of structural deformation reconstruction” section describes the FBG sensing system, theoretical foundation, and the proposed reconstruction framework; “Deformation reconstruction experiments and results” section presents experimental validation and results analysis; and “Conclusions” section provides concluding remarks. Mathematical modeling of structural deformation reconstruction Optical fiber sensing principle As one of the most popular optical fiber sensing technologies, FBG has been widely applied for structural condition monitoring due to its unique advantages over conventional electrical sensing technologies. The central wavelength shift of FBG (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}{\uplambda\:}$$\end{document}) is sensitive to strain and temperature changes. Assuming that the temperature remains constant, the strain-sensing principle of FBG could be expressed as24:1\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\frac{{\Delta\:}{\uplambda\:}}{{{\uplambda\:}}_{\text{B}}}=\left(1-{\text{P}}_{\text{e}}\right) \upvarepsilon \:\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\uplambda\:}}_{\text{B}}$$\end{document} is the initial central wavelength, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\varepsilon\:$$\end{document} is the strain change, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{e}}$$\end{document} is the effective photo-elastic coefficient of optical fiber. As the values of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{e}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\uplambda\:}}_{\text{B}}$$\end{document} are constant, there is a linear relationship between the strain change and the wavelength shift of FBG. FBG sensors could be used for quasi-distributed and multiple parameter measurements in hash environment due to their multiplexing capabilities such as time division multiplexing, wavelength division multiplexing, and space division multiplexing. Deformation reconstruction algorithms The strain change at measurement points on the upper and lower surfaces of a deformed component is depicted in Fig. 1. The relationship between the distance \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:(\text{v}$$\end{document}) from the neutral axis to the surface of the component and the component’s thickness\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left(\text{h}\right)$$\end{document}can be derived from the equivalent geometric relationship of strain change25:2\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}v=\left(\frac{{{\varepsilon\:}}_{\text{t}}}{{{\varepsilon\:}}_{\text{b}}-{{\varepsilon\:}}_{\text{t}}}\right)\ast\:h\end{array}$$\end{document} Fig. 1The relationship between thickness (h) and distance (v). The relationship between thickness (h) and distance (v). where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{\text{t}}$$\end{document} represents the strain at the measurement point on the component’s upper surface, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{\text{b}}$$\end{document} is the strain at the measurement point on the component’s lower surface as shown in Fig. 1. The primary focus of the research is on composite materials with constant cross-sections. For composite structures with constant cross-sections, strain can be converted into corresponding curvature based on the invariance of the material neutral layer and the relationship between curvature and radius of curvature:3\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\begin{array}{c}{\uprho\:}=\frac{2{\varepsilon\:}}{\text{h}}\end{array}\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\varepsilon\:}$$\end{document} is strain, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\uprho\:}$$\end{document} is curvature and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{h}$$\end{document} is the thickness of the structure. Slope recursion algorithm The structural deformation reconstruction method based on curvature transformation is to establish a relationship between surface strain and surface curvature to convert surface strain into discrete curvature. Subsequently, using the curvature information of the structural surface, a recursive algorithm and an interpolation algorithm are employed for point-to-line and line-to-surface modelling, thus reconstructing the three-dimensional shape of a surface. In this paper, the slope recursion algorithm is used for curvature transformation, and its schematic diagram is shown in Fig. 2. Fig. 2Schematic diagram of slope recursion algorithm. Schematic diagram of slope recursion algorithm. Curvature is the curve’s first and second-order derivative function, and slope is a first-order derivative function of the curve. Therefore, the curvature is a first-order derivative function of slope. The deduction of the coordinates at various corresponding positions on the curve can be obtained by calculating the slope from the curvature and the increment of coordinates of adjacent points on the curve based on the slope. In Fig. 2, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}+1}$$\end{document} are adjacent points located on the curve closely. The coordinates of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}+1}$$\end{document} are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({\text{x}}_{\text{n}},{\text{y}}_{\text{n}}\right)$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({\text{x}}_{\text{n}+1},{\text{y}}_{\text{n}+1}\right)$$\end{document}, respectively. Their corresponding curvatures are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\uprho\:}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\uprho\:}}_{\text{n}+1}$$\end{document}. The angles between the tangent and the x-axis are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\uptheta\:}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\uptheta\:}}_{\text{n}+1}$$\end{document}, while the corresponding slopes are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{k}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{k}}_{\text{n}+1}$$\end{document}. The change in the angles between the two points is \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}{{\uptheta\:}}_{\text{n}}$$\end{document}. According to the geometric relationship in the diagram, Eqs. (4) and (5) could be obtained:4\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\text{k}}_{\text{n}}= \text{tan}\left({{\uptheta\:}}_{\text{n}}\right),{\text{k}}_{\text{n}+1}=\text{tan}\left({{\uptheta\:}}_{\text{n}+1}\right)\:\:\:\:\:\:\end{array}$$\end{document}5\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{{\uprho\:}}_{\text{n}}=\frac{{\Delta\:}{{\uptheta\:}}_{\text{n}}}{{\Delta\:}{\text{S}}_{\text{n}}}=\frac{{{\uptheta\:}}_{\text{n}+1}-{{\uptheta\:}}_{\text{n}}}{{\Delta\:}{\text{S}}_{\text{n}}}\end{array}$$\end{document} From the above equation, the slope \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{k}}_{\text{n}+1}$$\end{document} can be calculated as:6\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\text{k}}_{\text{n}+1}=\text{tan}\left[{{\uprho\:}}_{\text{n}}\cdot\:{\Delta\:}{\text{S}}_{\text{n}}+\text{a}\text{r}\text{c}\text{t}\text{a}\text{n}\left({\text{k}}_{\text{n}}\right)\right]\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}{\text{S}}_{\text{n}}$$\end{document} is the arc length of segment \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}{\text{P}}_{\text{n}+1}$$\end{document}. The slope of each point can be obtained using Eq. (6) with the initial angle \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({{\uptheta\:}}_{1}\right)$$\end{document} and slope \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({\text{k}}_{1}\right)$$\end{document} of the curve. The increment of horizontal and vertical coordinates for each small interval can be calculated based on the slope:7\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\varDelta\: \text{x}=\frac{{\Delta\:}{\text{S}}_{\text{n}}}{\sqrt{1+{\text{k}}_{\text{n}}^{2}}}\end{array}$$\end{document}8\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\varDelta\: {\text{y}}={\text{k}}_{\text{n}}\cdot\:\varDelta\: {\text{x}}=\frac{{\text{k}}_{\text{n}}\cdot\:{\Delta\:}{\text{S}}_{\text{n}}}{\sqrt{1+{\text{k}}_{\text{n}}^{2}}}\end{array}$$\end{document} The recursive formulas for coordinates can be derived from Eqs. (7) and (8):9\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\text{x}}_{\text{n}+1}={\text{x}}_{\text{n}}+\varDelta\:x={\text{x}}_{\text{n}}+\frac{{\Delta\:}{\text{S}}_{\text{n}}}{\sqrt{1+{\text{k}}_{\text{n}}^{2}}}\end{array}$$\end{document}10\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\text{y}}_{\text{n}+1}={\text{y}}_{\text{n}}+\varDelta\:y={\text{y}}_{\text{n}}+\frac{{\text{k}}_{\text{n}}\cdot\:{\Delta\:}{\text{S}}_{\text{n}}}{\sqrt{1+{\text{k}}_{\text{n}}^{2}}}\end{array}$$\end{document} Ko displacement theory Ko displacement theory was initially proposed to monitor the deformation of aircraft wing structures during flight. It is based on classical material mechanics assumptions and the geometric displacement relationship of beam structure deformations, transforming the component deformation into a summation through segmentation and differential methods. According to the classical Euler-Bernoulli beam theory, the differential equation governing bending loads on a beam structure is:11\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\frac{{\text{d}}^{2}\text{y}}{\text{d}{\text{x}}^{2}}=\frac{\text{M}\left(\text{x}\right)}{\text{E}\text{I}}\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{y}$$\end{document} is deflection at a measurement point, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{x}$$\end{document} is the longitudinal coordinate of the tested component, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{M}\left(\text{x}\right)$$\end{document} is the bending moment on the tested component, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{E}$$\end{document} is the modulus of elasticity, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{I}$$\end{document} represents the moment of inertia of the beam section at position \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{x}$$\end{document}. The relationship between the bending moment \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left(\text{M}\left(\text{x}\right)\right)$$\end{document} on the tested component and the surface curvature strain (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\varepsilon\:}\left(\text{x}\right))$$\end{document} is:12\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\varepsilon\:\left(\text{x}\right)=\frac{\text{M}\left(\text{x}\right)\text{v}}{\text{E}\text{I}}\end{array}$$\end{document} By substituting Eq. (12) into Eq. (11), it can be concluded that:13\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\frac{{\text{d}}^{2}\text{y}}{\text{d}{\text{x}}^{2}}=\frac{{\varepsilon\:}\left(\text{x}\right)}{\text{v}}\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{v}$$\end{document} is the distance from the neutral axis of the cross-section to the beam surface. A beam structure can be discretized as a continuous body based on a discrete approach. In the beam’s lengthwise direction, it is divided into small elements. Assuming that the elements are linearly distributed and the beam with a length of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{l}$$\end{document} is divided into n discrete elements the with lengths of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}\text{l}$$\end{document}, the coordinates along the length direction of the first node and last node of t segment of the element are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{i}-1}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{i}}$$\end{document}, respectively. The corresponding strains of the two nodes are \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{\text{i}-1}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{\text{i}}$$\end{document}. The strain expression on this element can be obtained through linear fitting method:14\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\varepsilon\:\left(\text{x}\right)={{\varepsilon\:}}_{\text{i}-1}+\frac{{{\varepsilon\:}}_{\text{i}}-{{\varepsilon\:}}_{\text{i}-1}}{{\Delta\:}\text{l}}\:\left(\text{x}-{\text{x}}_{\text{i}-1}\right),\:\varDelta\:l=\frac{\text{l}}{\text{n}}\:\end{array}$$\end{document} For an element in the range of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{i}-1}<\text{x}<{\text{x}}_{\text{i}}$$\end{document}, the slope \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left(\text{t}\text{a}\text{n}{\uptheta\:}\left(\text{x}\right)\right)$$\end{document} can be obtained by integrating Eq. (13):15\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\text{tan}\uptheta\:\left(\text{x}\right)=\int\limits_{{\text{x}}_{\text{i}-1}}^{\text{x}}\frac{{\text{d}}^{2}\text{y}}{\text{d}{\text{x}}^{2}} \text{dx}+\text{tan}{{\uptheta\:}}_{\text{i}-1}\:\:\:{\text{x}}_{\text{i}-1}\le\:\text{x}\le\:{\text{x}}_{\text{i}}\end{array}$$\end{document} The displacement \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\text{y}\left(\text{x}\right)$$\end{document} in the range of \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{i}-1}\le\:\text{x}\le\:{\text{x}}_{\text{i}}$$\end{document} is composed of the integral of slope and the displacement:16\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c} \text{y}\left(\text{x}\right)=\int\limits_{{\text{x}}_{\text{i}-1}}^{\text{x}}\text{t}\text{a}\text{n}{\uptheta\:}\left(\text{x}\right) \text{dx}+{\text{y}}_{\text{i}-1}\:\:\:\:\:\:\:{\text{x}}_{\text{i}-1}\le\:\text{x}\le\:{\text{x}}_{\text{i}}\end{array}$$\end{document} Substituting Eq. (14) into Eqs. (15) and (16), the incremental slope \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({\Delta\:}\text{t}\text{a}\text{n}{{\uptheta\:}}_{\text{i}}\right)$$\end{document} and incremental displacement \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({{\Delta\:}\text{y}}_{\text{i}}\right)$$\end{document} at position \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{i}}$$\end{document} can be obtained:17\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\varDelta\:\text{tan}{{\uptheta\:}}_{\text{i}}=\frac{{\Delta\:}\text{l}}{2\text{v}}\left({{\varepsilon\:}}_{\text{i}-1}+{{\varepsilon\:}}_{\text{i}}\right)\end{array}$$\end{document}18\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}{\text{y}}_{\text{i}}=\frac{{\left({\Delta\:}\text{l}\right)}^{2}}{6\text{v}}\left(2{{\varepsilon\:}}_{\text{i}-1}+{{\varepsilon\:}}_{\text{i}}\right)+{\Delta\:}\text{l}\ast\:\text{t}\text{a}\text{n}{{\uptheta\:}}_{\text{i}-1}\:\:\:\:\:\:$$\end{document} As it is difficult to determine the distance \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:(\text{v}\:\text{i}\text{n}\:\text{E}\text{q}.\:(2\left)\right)$$\end{document} from the neutral axis to the surface of the component’s cross-section, accurately, the reconstruction Eq. (17) should be optimized. The optimized Eqs. (19) and (20) for the deformation reconstruction of the component can be obtained by substituting Eq. (2) into Eqs (17) and (18).19\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\varDelta\:\text{tan}{{\uptheta\:}}_{\text{i}}=\frac{{\Delta\:}\text{l}}{2\text{h}}\left({\left({{\varepsilon\:}}_{\text{b}}-{{\varepsilon\:}}_{\text{t}}\right)}_{\text{i}-1}+{\left({{\varepsilon\:}}_{\text{b}}-{{\varepsilon\:}}_{\text{t}}\right)}_{\text{i}}\right)\:\:\:\end{array}$$\end{document}20\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}{\text{y}}_{\text{i}}=\frac{{\left({\Delta\:}\text{l}\right)}^{2}}{6\text{h}}\left(2{\left({{\varepsilon\:}}_{\text{b}}-{{\varepsilon\:}}_{\text{t}}\right)}_{\text{i}-1}+{\left({{\varepsilon\:}}_{\text{b}}-{{\varepsilon\:}}_{\text{t}}\right)}_{\text{i}}\right)+{\Delta\:}\text{l}\text{t}\text{a}\text{n}{{\uptheta\:}}_{\text{i}-1}$$\end{document} The overall process of the reconstruction algorithm is shown in Fig. 3. Fig. 3The overall process of the reconstruction algorithm. The overall process of the reconstruction algorithm. Algorithm optimization Optimization of slope recursion algorithm In the slope recursion algorithm, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}+1}$$\end{document} are sharp corner points to construct a right triangle through the arc length \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\left({\Delta\:}{\text{S}}_{\text{n}}\right)$$\end{document} approximate bevel length. The displacement increment is derived from the trigonometric function relationship \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}\text{x}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\Delta\:}\text{y}$$\end{document}. The tangent value of the trigonometric function is calculated by replacing the slope at point \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}$$\end{document} on the curve. The curvature on the curve is continuously changing. When the micro-arc segments are taken for reconstruction and the slopes of one side points are used for approximate replacements, the reconstruction error of each micro-arc segment changes in the same direction, and the overall error will be cumulated. The acute angle at point \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}$$\end{document} is located in the interval formed by the tangent values at point \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{n}+1}$$\end{document}. The overall cumulative error can be reduced by substituting the mean value of the two tangent angles into the tangent function. The displacement increments in X and Y directions can be calculated by Eqs. (21) and (22)21\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\text{x}}_{\text{n}+1}={\text{x}}_{\text{n}}+\varDelta\:x={\text{x}}_{\text{n}}+\frac{{\Delta\:}{\text{S}}_{\text{n}}}{\sqrt{1+{\text{t}\text{a}\text{n}}^{2}\left(\frac{{{\uptheta\:}}_{\text{n}}+{{\uptheta\:}}_{\text{n}+1}}{2}\right)}}\end{array}$$\end{document}22\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{\text{y}}_{\text{n}+1}={\text{y}}_{\text{n}}+\varDelta\:y={\text{y}}_{\text{n}}+\frac{\text{t}\text{a}\text{n}\left(\frac{{{\uptheta\:}}_{\text{n}}+{{\uptheta\:}}_{\text{n}+1}}{2}\right)\cdot\:{\Delta\:}{\text{S}}_{\text{n}}}{\sqrt{1+{\text{t}\text{a}\text{n}}^{2}\left(\frac{{{\uptheta\:}}_{\text{n}}+{{\uptheta\:}}_{\text{n}+1}}{2}\right)}}\end{array}$$\end{document} Fitting of strain data based on cross validation The structural deformation reconstruction methods using the slope recursion algorithm and Ko displacement theory are both based on discrete concepts. The number of specific sensor placements within a certain range is positively correlated with reconstruction accuracy. Considering the cost factors in engineering, it’s challenging to deploy a large number of sensors. It is difficult to obtain the strain at structural boundaries using optical fiber sensors, which involves data expansion and complementation of boundary strain through fitting or interpolation. However, there is a lack of universality for the fitting of strain data due to the complexity of the data and model. The minimum error strain fitting method based on cross-validation can be proposed. The estimated strain at the fixed end can be calculated through the root mean square error between the measured strain data and the finite element simulation data. The strain at the fixed end with the minimum calculation error can be expressed as:23\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}{{\varepsilon\:}}_{0}={{\varepsilon\:}}_{0}^{{\prime\:}}+ \text{sign}\left({\sum\:}_{\text{i}=1}^{\text{n}}({{\varepsilon\:}}_{\text{i}}-{{\varepsilon\:}}_{\text{i}}^{{\prime\:}})\right)\ast\:\sqrt{\frac{{\sum\limits}_{\text{i}=1}^{\text{n}}({{\varepsilon\:}}_{\text{i}}-{{\varepsilon\:}}_{\text{i}}^{{\prime\:}}{)}^{2}}{\text{n}}}\:\:\:\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{\text{i}}(\text{i}=\text{1,2},\dots\:,\text{n})$$\end{document} is the measured strain using sensors, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{\text{i}}^{{\prime\:}}(\text{i}=\text{0,2},\dots\:,\text{n})$$\end{document} is the strain obtained by finite element simulation, and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\varepsilon\:}}_{0}$$\end{document} is the calculated strain at the fixed end. The cross-validation of the fitting result will be carried out. A fusion dataset will be constructed by using the calculated strain at the fixed support end and the measured strain. A small portion of the measured strain will be randomly selected as the test sets, and the rest will be used to build the least squares polynomial fitting function. The fitting data will be compared with the corresponding test data, using the root mean square error as a quantification standard. Multiple sets of fitting results will be obtained through iterative cross-validation, getting the final fitting result with minimum error. Figure 4 shows the flowchart of the strain data fitting algorithm based on cross-validation. Among them, i represents the number of cross-validation iterations.i₂ represents the maximum degree of the fitted polynomial constructed in a single run.M represents the number of data points used for constructing the fitting curve. Fig. 4Flowchart of strain data fitting algorithm based on cross validation. Flowchart of strain data fitting algorithm based on cross validation. The two-terminal recursive weighted reconstruction method The slope recursive reconstruction algorithm, when used to calculate structures fixed at both ends, results in cumulative end errors that cause discrepancies between the reconstructed and actual results. A two-terminal recursive weighted reconstruction method is proposed based on the initial conditions of the structure. Reconstructions are performed from both fixed ends as starting points, and the weighted sum of reconstructions is calculated using the inverse distance weighting method. Results closer to the starting point should have a larger weight, while those further from the starting point should have a smaller weight. The inverse distance weighting combination formula is:24\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}\begin{array}{c}{\text{z}}_{\text{i}}=\left(1-\frac{\left|{\text{x}}_{\text{i}}-{\text{x}}_{\text{L}}\right|}{\left|{\text{x}}_{\text{R}}-{\text{x}}_{\text{L}}\right|}\right){\text{z}}_{\text{L}\text{i}}+\left(1-\frac{\left|{\text{x}}_{\text{i}}-{\text{x}}_{\text{R}}\right|}{\left|{\text{x}}_{\text{R}}-{\text{x}}_{\text{L}}\right|}\right){\text{z}}_{\text{R}\text{i}}\:\end{array}\end{array}$$\end{document} where, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{L}}$$\end{document} and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{R}}$$\end{document} are X-axis coordinates of the start point and the endpoint, respectively. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{x}}_{\text{i}}$$\end{document} is the X-axis coordinate of the point. \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{z}}_{\text{L}\text{i}}$$\end{document} is the reconstruction value from this point to the starting point and \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{z}}_{\text{R}\text{i}}$$\end{document} represents the reconstruction value of the point starting from the endpoint. Deformation reconstruction experiments and results Reconstruction experiments of a composite plate The deformation results of a composite plate using the reconstruction algorithms were compared with the results obtained by finite element simulation analysis. Optical fiber sensors can be employed to detect multi-point strain on the composite plate. The deformation reconstruction experiments were carried out to validate the proposed deformation reconstruction methods. Root Mean Square Error (RMSE) was used to evaluate the deformation reconstruction accuracy of the proposed methods, which is defined as:25\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:\begin{array}{c}E=\sqrt{\frac{{\sum\:}_{\text{i}=1}^{\text{n}}{\text{e}}_{\text{i}}^{2}}{\text{n}}}=\sqrt{\frac{{\sum\:}_{\text{i}=1}^{\text{n}}{({\text{d}}_{\text{i}}-{{\text{d}}_{\text{i}}}^{{\prime\:}})}^{2}}{\text{n}}}\:\:\:\end{array}$$\end{document} where \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{d}}_{\text{i}}$$\end{document} is the reference displacement at point \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{i}}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{{\text{d}}_{\text{i}}}^{{\prime\:}}$$\end{document} is the reconstructed displacement at point \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{P}}_{\text{i}}$$\end{document}, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\:{\text{e}}_{\text{i}}$$\end{document} represents the absolute error of the reconstructed displacement, and n represents the number of reconstructed points. The overall reconstruction accuracy of η can be defined as:26\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$\eta = \frac{{\mathop \sum \nolimits_{{{\text{i}} = 1}}^{{\text{n}}} {{\upeta }}_{{\text{i}}} }}{{\text{n}}} = \frac{{\mathop \sum \nolimits_{{{\text{i}} = 1}}^{{\text{n}}} 1 - \left| {\frac{{{\text{d}}_{{\text{i}}} - {\text{d}}_{{\text{i}}}^{\prime } }}{{{\text{d}}_{{\text{i}}} }}} \right| \cdot 100{\text{\% }}}}{{\text{n}}}$$\end{document} Figure 5 shows the installation of FBG sensors on the carbon fiber composite plate. The length, width and thickness of the plate are 480 mm, 400 mm and 5 mm, respectively. One end of the plate was fixed. 20 FBG sensors in 4 strings (CH1, CH2, CH3 and CH4) were glued on the surface of the plate to measure the strain. The wavelengths of 5 FBG sensors in the same are from 1530 nm to 1550 nm with an interval of 5 nm, and the installation distance is 80 mm. The distance between FBG strings is 80 mm. Fig. 5Installations of FBG sensors on the composite plate (Units: mm). Installations of FBG sensors on the composite plate (Units: mm). Finite element simulation analysis A 3D model of the composite plate was established with the material of epoxy carbon woven. The strain distribution and deformation displacement of the plate under loads were calculated using the software of ANSYS (Ansys workbench 2024 R2, URL: https://www.ansys.com/). The six-node element SOLID186 was chosen for meshing. The physical environment was set to Mechanical, and the structure was divided into hexahedral mesh types using the Multizone method with a mesh size of 5 mm. The plate model was divided into 8,204 elements with a total of 58,336 nodes. The end of the plate along the length direction is fixed. 4 solving paths at the same positions with FBG sensor strings were set on the surface of the plate. There were 49 solving points of strain and displacement on each path. A force of 40 N was applied on the plate at a corner of the cantilever end. The strain and displacement at the solving points were calculated. The reconstruction method based on the slope recursion algorithm can be improved by using tangent angle mean values. The Ko displacement theory can be improved by cross-validating the fitting of strain data. The simulated strain data on the surface of the composite plate were employed as inputs to reconstruct the deformation using the proposed methods. Table 1 lists the reconstruction accuracies of displacements using the proposed algorithms compared with the results obtained by simulation. The RMSE of the improved slope recursion algorithm decreases from 0.1193 to 0.0153. The reconstruction accuracy using the improved slope recursion algorithm is increased from 94.61% to 99.03%. The reconstruction method using the improved Ko displacement theory has the lowest RMSE of 0.0117. The reconstruction accuracy using improved Ko displacement is 99.41%, which is higher than the accuracy of 99.04% using Ko displacement theory. The overall deformation results of the plate obtained by simulation and reconstruction using the improved slope recursion algorithm are shown in Fig. 6. Fig. 6Deformations of the composite plate: (a) simulation and (b) reconstruction using the improved slope recursion algorithm. Deformations of the composite plate: (a) simulation and (b) reconstruction using the improved slope recursion algorithm. Table 1Deformation reconstruction accuracy of the composite plate under a force at a cantilever corner.Reconstruction algorithmSlope recursion algorithmImproved slope recursion algorithmKo displacement theoryImproved Ko displacement theoryRMSE0.11930.01530.01600.0117Overall accuracy94.61%99.03%99.04%99.41% Table 1Deformation reconstruction accuracy of the composite plate under a force at a cantilever corner.Reconstruction algorithmSlope recursion algorithmImproved slope recursion algorithmKo displacement theoryImproved Ko displacement theoryRMSE0.11930.01530.01600.0117Overall accuracy94.61%99.03%99.04%99.41% Deformation reconstruction accuracy of the composite plate under a force at a cantilever corner. The comparisons between the reconstructed displacements on the solving path of CH3 with the results obtained by simulation are shown in Fig. 7. The maximum displacement occurs at the cantilever end of the plate, which is shown in an enlarged view. Figure 7a simulates the reconstruction results of Ko displacement theory under different sensor numbers. As the number of sensors increases, the reconstruction accuracy improves. The reconstruction result with 17 strain points as inputs has the highest accuracy, with a relative error of 0.55% at the endpoints. Figure 7b reflects the reconstruction results of different algorithms with 5 strain points as inputs. Compared with the displacement obtained by simulation, the relative error of the reconstructed displacement calculated by the slope recursion algorithm, the improved recursion algorithm, the Ko displacement theory and the improved Ko displacement theory are 2.49%, 0.7%, 0.75% and 0.51%, respectively. The reconstruction method using the improved Ko displacement theory has the highest accuracy. Fig. 7Comparison of reconstruction results on the solving path of CH3: (a) different numbers of sensors and (b) different reconstruction algorithms. Comparison of reconstruction results on the solving path of CH3: (a) different numbers of sensors and (b) different reconstruction algorithms. Table 2 shows the reconstructed accuracies using the proposed algorithms when the force of 40 N was applied at the middle edge of the plate cantilever end. The deformations of the plate obtained by simulation and the reconstruction method based on the improved Ko displacement theory are shown in Fig. 8a and b, respectively. The RMSE using the improved slope recursion algorithm decreases from 0.1054 to 0.0105. The deformation reconstruction accuracy of the improved slope recursion algorithm is increased to 99.23%, which is higher than the results using the slope recursion algorithm. The reconstruction method using the improved Ko displacement theory has the highest accuracy of 99.71%, which is slightly increased compared with the results using the Ko displacement theory. Fig. 8Deformations of the composite plate: (a) simulation and (b) reconstruction using the improved Ko displacement theory. Deformations of the composite plate: (a) simulation and (b) reconstruction using the improved Ko displacement theory. Table 2Reconstruction accuracy of the composite plate under a load at the middle edge of the cantilever end.Reconstruction algorithmSlope recursion algorithmImproved slope recursion algorithmKo displacement theoryImproved Ko displacement theoryRMSE0.10540.01050.00610.0074Overall accuracy94.70%99.23%99.63%99.71% Table 2Reconstruction accuracy of the composite plate under a load at the middle edge of the cantilever end.Reconstruction algorithmSlope recursion algorithmImproved slope recursion algorithmKo displacement theoryImproved Ko displacement theoryRMSE0.10540.01050.00610.0074Overall accuracy94.70%99.23%99.63%99.71% Reconstruction accuracy of the composite plate under a load at the middle edge of the cantilever end. Experiments and results To verify the proposed deformation reconstruction methods, deformation reconstruction experiments of a composite plate were carried out using the setup shown in Fig. 9. One end of the plate was fixed, and the other end was free. To ensure the accuracy of the reconstruction algorithm, the strain measurement system must be equipped with high accuracy and stability. The FBG sensors require an extensive strain tolerance ( ≥ ± 5000 µε), high accuracy (± 1–5 µε), and sub-microstrain resolution to ensure reconstruction fidelity. Strategic wavelength spacing (> 3–5 nm) prevents spectral overlap. The interrogator must provide multi-channel synchronous acquisition, a broad optical spectrum (≥ 80 nm), and pm-level resolution, with sampling rates adapted to static (≥ 1 Hz) or dynamic (≥ 100 Hz) conditions. 20 FBG sensors in 4 strings were installed on the surface of the plate at the positions shown in Fig. 5 to measure strain. An FBG interrogator (type: HI-8000-16, resolustion: 0.1 pm, accuracy: ±1pm, channel: 16, sampling rate: 50 Hz) and a computer was used to monitor the wavelengths of FBG sensors. A laser displacement sensor with a type of MSE-TS803 and a linear accuracy of ± 0.05% F.S. was employed to detect the deformation displacement of the plate under loads. Fig. 9Experimental setup of a composite plate. Experimental setup of a composite plate. Deformation experiments of the plate under different forces and locations were conducted. The forces of 30 N and 40 N at the corner or middle edge of the cantilever end were generated by several weights. The multi-point strain on the surface of the plate measured by FBG sensors was employed for the deformation reconstruction using the proposed reconstruction methods. Figure 10 illustrates the reconstructed deformation of the plate under different loads using the improved slope recursion algorithm. Fig. 10Reconstruction results based on the improved slope recursion algorithm under different loads. Reconstruction results based on the improved slope recursion algorithm under different loads. The reconstructed results were compared with the measurements from the laser displacement sensor. Table 3 lists the overall reconstruction accuracy calculated using Eq. (25). The average overall reconstruction accuracy calculated by the slope recursion algorithm, the improved slope recursion algorithm, the Ko displacement theory and the improved Ko displacement theory are 89.04%, 90.95%, 90.71% and 91.63%, respectively. The overall reconstruction accuracy using the improved Ko displacement theory is highest compared with the other algorithms. The reconstruction accuracies using the algorithms of slope recursion algorithm and the Ko displacement theory increase after improvements. The physical interpretation here is, in the derivation process, the slope recurrence method uses partial approximate substitutions, resulting in lower single-segment reconstruction accuracy than the Ko displacement theory, and the overall reconstruction process also produces more cumulative errors. In contrast, the improved Ko displacement theory can obtain more accurate strain fitting curves compared with the unimproved version. Table 3Overall deformation reconstruction accuracies under different loads.LoadsSlope recursion algorithm18 (%)Improved slope recursion algorithm (%)Ko displacement theory16 (%)Improved Ko displacement theory (%)30 N at corner86.7487.2587.2287.3940 N at corner90.7292.0091.7693.2930 N at middle edge84.9789.0988.4190.3940 N at middle edge93.7295.4595.4695.46 Table 3Overall deformation reconstruction accuracies under different loads.LoadsSlope recursion algorithm18 (%)Improved slope recursion algorithm (%)Ko displacement theory16 (%)Improved Ko displacement theory (%)30 N at corner86.7487.2587.2287.3940 N at corner90.7292.0091.7693.2930 N at middle edge84.9789.0988.4190.3940 N at middle edge93.7295.4595.4695.46 Overall deformation reconstruction accuracies under different loads. Reconstruction experiments of a composite plate Composite material reinforced structures have been widely implemented to increase strength and stiffness. Reconstruction experiments of a reinforced structure of carbon fiber composite material were carried out to validate the proposed reconstruction methods. As shown in Fig. 11, the dimensions of the base plate on the structure are 385 mm long, 350 mm wide and 1.7 mm thick, which has a square hole of 80 mm × 80 mm at the centre. The reinforcement structure consists of 2 transverse rib plates and 2 longitudinal rib plates, affixed by bolts. Fig. 11Simplified 3D model of a reinforced structure. Simplified 3D model of a reinforced structure. Finite element simulation analysis The strain distribution and deformation displacement of the structure under loads were also calculated using the software of ANSYS as previously stated. The physical environment was set to Mechanical. The structure was divided using the tetrahedrons method. The patch conforming algorithm was chosen for the division. The reinforcement structure was divided into tetrahedral meshes with a unit size of 2 mm, resulting in a simulation model with a total of 1,171,536 nodes and 696,337 elements. 7 strain solving paths with 200 solving points on each on the surface of the base plate were set along X direction. The two ends of the structure along length direction were fixed, and loads were applied in the normal direction of the plate. The boundary constraints and deformation results with a force of 5kN are shown in Fig. 12a and b, respectively. Fig. 12Finite element simulation of a reinforced structure: (a) boundary constraints and (b) deformation solution results in the Z direction . Finite element simulation of a reinforced structure: (a) boundary constraints and (b) deformation solution results in the Z direction . The strain distributions are symmetric due to the symmetric structure. The displacements at the fixed end of the structure are nearly zero. Significant deformation is observed where there is no rib support. Figure 13 shows the comparison of deformation reconstruction results along different solving paths. On the first solving path, the RMSE of the reconstruction results for the slope recursive algorithm, Ko displacement theory, and the two-terminal recursive reconstruction method are 0.5186 mm, 0.8517 mm and 0.4014 mm, respectively. On the second solving path, they are 0.0740 mm, 0.1685 mm and 0.0335 mm, respectively. The reconstruction error of the two-terminal recursive weighted reconstruction method is significantly lower than that of other methods. The strain on the surfaces of the base plate calculated by simulation was employed as inputs for the deformation reconstructions of the proposed reconstruction algorithms. Fitting methods are not used to process the strain data during the reconstruction process. Figure 14a and b respectively display the deformation obtained from the simulation solution and the deformation reconstruction result in the Z-direction solved by the two-terminal recursive weighted reconstruction method. The overall reconstruction accuracy of the two-terminal recursive weighted reconstruction method is 88.67%. The error in the reconstruction results is primarily concentrated at the location of maximum deformation of the plate. Fig. 13Comparison of reconstruction results on the solving path under a force of 5kN: (a) reconstruction results on the first solving path (Fig. 16b Col.1) and (b) reconstruction results on the second solving path(Fig. 16b) Col.2). Comparison of reconstruction results on the solving path under a force of 5kN: (a) reconstruction results on the first solving path (Fig. 16b Col.1) and (b) reconstruction results on the second solving path(Fig. 16b) Col.2). Fig. 14Deformation reconstructions of the reinforced composite structure: (a) Simulation results and (b) reconstructed results using the two-terminal recursive weighted reconstruction method. Deformation reconstructions of the reinforced composite structure: (a) Simulation results and (b) reconstructed results using the two-terminal recursive weighted reconstruction method. Experiments and results To validate the proposed reconstruction algorithms further, deformation reconstruction experiments of a composite material reinforced structure were conducted. As shown in Fig. 15, the experimental setup consists of a composite material reinforcement structure, a material testing machine (Instron 3382), FBG sensors, an FBG interrogator, a PC, and optical fiber cables. Fig. 15Experimental setup of a composite material reinforced structure. Experimental setup of a composite material reinforced structure. 70 FBG sensors in several strings were installed by epoxy resin adhesive on the surfaces of the reinforced structure to measure the strain distribution under loads. Figure 16 shows the installation locations on the structure. The lower end of the material testing machine was fixed, and a vertical pressure was applied downward at the upper end. 20 FBG sensors and 2 FBG sensors are placed on the surface of the structure with rib plates to detect strain in the X direction and Y direction, respectively, as shown in Fig. 16a. 48 FBG sensors were arranged along the X direction in 7 rows and 7 columns on the surface of the structure without rib plates, as shown in Fig. 16b. Fig. 16Installations of FBG sensors on the reinforced composite structure: (a) FBG sensors on the surface with rib plates and (b) FBG sensors on the surface without rib plates. Installations of FBG sensors on the reinforced composite structure: (a) FBG sensors on the surface with rib plates and (b) FBG sensors on the surface without rib plates. Figure 17 shows the strain measured by FBG sensors in X direction on the surface of the structure without rib plates under a compression force of 20kN. Based on the strain data, the constructed strain distribution on the surface is shown in Fig. 18a. Figure 18b shows the strain distribution by the finite element analysis data under identical loading conditions. The overall strain distribution is symmetrical, and similarities are observed between the two. Fig. 17Strain measured by FBG sensors in X direction. Strain measured by FBG sensors in X direction. Fig. 18Reconstructed strain distribution in X direction: (a) reconstructed strain distribution by measured strain and (b) reconstructed strain distribution by simulated strain. Reconstructed strain distribution in X direction: (a) reconstructed strain distribution by measured strain and (b) reconstructed strain distribution by simulated strain. In simulation, one end of the structure is fixed, and the force is applied at the other end. However, the boundary conditions in the experiment are different as both ends of the structure are subjected to the load. Both the upper and lower ends of the structure produced displacements. The two-terminal recursive reconstruction method was used for the deformation reconstruction of the structure. The abnormal measured strain data caused by poor installations were corrected according to the simulated data. The simulation data were proportionally converted and added to the measured strain data. The piecewise linear interpolation method was used to supplement the data. Figure 19a and b show the reconstructed deformation of the structure using the slope recursion algorithm and the two-terminal recursive weighted reconstruction method, respectively. The displacements at positions with rib plates are much smaller than those at positions without rib plates. The strength and stiffness of the reinforced structure increase compared with the structure without rib plates. Fig. 19Reconstructed deformations of the reinforced composite plate: (a) the slope recursion algorithm and (b) the two-terminal recursive reconstruction method. Reconstructed deformations of the reinforced composite plate: (a) the slope recursion algorithm and (b) the two-terminal recursive reconstruction method. Conclusions Structural health monitoring is essential for composite material structures for damage assessment and predictive maintenance to prevent potential safety risks. Deformation monitoring based on optical fiber sensing technology plays an important role in structural health monitoring. In this paper, deformation reconstruction methods of composite material structures based on optical fiber sensors have been investigated by improving the slope recursion algorithm and Ko displacement theory. The method of angle averaging was employed to optimize the slope recursion algorithm, solving the problem of accumulating errors caused by single-sided slope approximation. The cross-validation method with finite element simulated data was used to obtain the fitting strain data with smaller errors and improve the reconstruction accuracy of the Ko displacement theory. Numerical analysis and experimental investigation on a composite plate and a composite material reinforced structure were conducted to validate the effectiveness of the improved slope recursion algorithm and the improved Ko displacement theory, which was validated through finite element simulation of composite material structures. FBG sensors were installed on the carbon fiber composite plate to measure the strain under deformation. The experimental results of structural deformation reconstruction demonstrate that the improved reconstruction algorithms achieve higher reconstruction accuracy compared with the initial algorithms. Compression experiments were carried out on the reinforced structure. The results of compression experiments on the reinforced structure indicate the feasibility of the proposed deformation reconstruction method for complex structures. It is important to note that the algorithms presented in this paper were validated only under single-point loading conditions. Their robustness under more complex multi-point or random loads, which are common in practical applications, remains a subject for future investigation and represents a key limitation of the current study. Further research will focus on the sensor layout scheme for complex composite structures and the analysis of reconstruction accuracy based on the measurement of actual deformation results. Additionally, the relationship between high-resolution strain measurement using optical frequency domain reflectometry (OFDR) and deformation reconstruction accuracy will be a key area of study. The temperature compensation methods of FBG sensors will be investigated for the structural condition monitoring of composite material structures.