Liquid Level Fiber Sensor Based on Fully Connected Neural Network and In-line Mach-Zehnder Interferometer

An innovative fiber optic liquid level sensor based on a peanut-shaped fiber structure has been designed and fabricated for use in micro-buoy systems. This sensor features two cascaded peanut-shaped structures with a 25 mm spacing, and its sensing characteristics have been thoroughly studied. This setup achieves millimeter-level monitoring intervals by measuring changes in interference wavelengths, enabling the system to accurately deduce the current liquid level state with a sensitivity of 0.087 $\mathrm{nm} / \mathrm{mm}$. To enhance the accuracy of data analysis, a Fully Connected Neural Network (FCNN) has been implemented for data processing and analysis. After training this model, the average percentage error was successfully reduced to 0.76%. Consequently, this cutting-edge fiber optic liquid level sensor demonstrates significant potential for precise measurements in marine micro-buoy systems, thereby ensuring seamless monitoring and robust protection of offshore wind energy installations and critical port infrastructures. The design effectively demonstrates the potential development space for the application of artificial intelligence and fiber optic sensing technology in oceanic micro-buoy liquid level measurement.