Smart Sounding Table Using Adaptive Neuro-Fuzzy Inference System

Abstract

Marine engineers measure the liquid level (sounding depth) to calculate the volumetric content of a ship's tank. The sounding depth is determined using an ullage pipe located at specific points on the tanks. To estimate the accurate volume of liquid, considering the ship's trim and heel conditions, engineers use a tank table (sounding table) consisting of hundreds of pages. However, this method is time-consuming and lacks intermediate values for sounding depth, trim, and heel. Ship designers recommend to use linear interpolation for intermediate values, yet this process is also time-consuming. This paper proposes the implementation of an Adaptive Neuro-Fuzzy Inference System (ANFIS) to digitize the sounding table. To our knowledge, we are the first to apply the ANFIS method to develop a model for liquid volume in non-uniform geometric tanks, accounting for different trim and heel conditions of the vessel. In this study, the digitization of the sounding table using ANFIS is referred to as the Smart Sounding Table (SST). SST's accuracy is validated against experimental values, revealing an R-squared value of 0.9999, a mean absolute percentage error of 0.3515, and a root mean square error of 0.0366. These metrics clearly show that the SST algorithm accurately and reliably models experimental data. Marine engineers input three parameters (sounding depth, trim, and heel) into the SST, enabling rapid and accurate determination of liquid volume in their tanks, without the need for interpolation or exhaustive page searches.