A reflective multimode fiber vector bending sensor based on specklegram

The vector bending sensor that senses bend orientation and curvature is desperately needed in the smart medical field. However, the existing fiber-optic vector bending sensor is challenging to detect the 360 degrees bending orien-tation, and both sides of the transmission structure need related equipment resulting in limited application. Here, a reflective multimode fiber vector bending (RMVB) sensor based on the specklegrams is proposed to overcome these challenges. Light source coupling and specklegrams capturing at the same side are achieved by the beamsplitter prism at the proximal end and the reflective coating at the distal end. Specklegrams can be used to estimate the bending state since the bending alters the mode interference. The bending state prediction network is trained using a dataset of 360 degrees bending orientations and 0-5.7 m-1 curvatures. The bending orientation and curvature interval are predicted by the classification model with accuracies of 0.99 and 0.97, respectively. With the mean absolute error lower than 4 % of the measurement range, the regression model can accurately predict the bending orientation and curvature. Therefore, the RMVB sensor with a novel structure capable of curvature sensing in 360 degrees bending orientation provides a promising approach for a single-fiber vector bending sensor and is expected for application in space-tight applications such as interventional operation.