Search

• This study presents an error analysis and correction method for multi-core fiber to improve the accuracy of shape sensing. • Monte Carlo method was used to simulate and analyze the sensing error. • An MLP neural network is built to reduce the sensing error, which is suitable for multi-core fiber with different structures. • The sensing experiments under different curvature confirmed that method improves the sensing accuracy of multi-core fiber. To meet the application requirements of accurate shape sensing for biomedical robotics and flexible morphing structure of aircraft etc, the error analysis and correction method for multi-core fiber is proposed. First, the relationship between the central wavelength shift and curvature of the multi-core fiber is analyzed which is based on the principle of multi-core fiber shape sensing, and a temperature decoupling method was researched. Considering the influence of the fiber structural parameters on curvature sensing, which is simulated by using the Monte Carlo method, and the simulation results show a strong influence among them. Next, to reduce the sensing errors introduced by the structural parameters, an error correction method based on regularized multilayer perceptron (MLP) networks is investigated, the proposed method does not require precise structural parameters of the multi-core fiber. Finally, some experiments with different curvatures are carried out, the results showed that the maximum root mean square error (RMSE) decreases from 2.03 m−1 to 0.007 m−1, and the minimum RMSE decreases from 0.01 m−1 to 0.001 m−1. The method this paper proposed has potential applications in shape sensing in the fields of biomedical engineering and aerospace. [ABSTRACT FROM AUTHOR]