Data-driven multi-joint waveguide bending sensor based on time series neural network.

Area of Science:

• Optoelectronics and Sensor Technology
• Biomedical Engineering
• Materials Science

Background:

• Traditional fiber optic sensing systems are limited by bulky equipment and wired connections, restricting their use to large-scale facilities.
• Advancements in neural networks and flexible materials enable new applications, particularly in bionics.
• There is a need for compact, flexible sensors capable of detailed motion detection for integration into biological systems.

Purpose of the Study:

• To design and fabricate a multi-joint waveguide bending sensor using color-dyed filters for detecting bending.
• To characterize the optical properties of the sensor materials to understand the underlying principles.
• To quantitatively predict bending position and angle using time series neural networks.

Main Methods:

• Fabrication of the waveguide bending sensor using a casting method with soft silicone rubber.
• Characterization of optical properties of the sensor materials.
• Application of time series neural networks for quantitative prediction of bending parameters.

Main Results:

• The developed waveguide bending sensor successfully detects bending angles, directions, and positions.
• The optical properties of the silicone rubber and color-dyed filters were characterized.
• Neural network analysis provided accurate quantitative predictions of bending parameters.

Conclusions:

• The multi-joint waveguide bending sensor demonstrates effective performance for detecting complex bending motions.
• The data-driven neural network demodulation approach shows significant promise for engineering applications.
• This flexible sensor technology opens new avenues for wearable devices and bionic interfaces.