Cake Moisture Estimation Based on Image Analysis and Regression Model for Controlling the Compression Time of Filter Press in Sludge Dewatering

Abstract

This study proposes practical methods for estimating the moisture content of sludge, represented by the cake moisture, in the filter press dewatering process. Because the cake moisture and filtrate volume are difficult to measure directly, the proposed approaches utilize indirectly measurable data, including drain outlet images and the differential pressure during the compression phase. By analyzing the correlations between these parameters and the cake moisture, estimation models were developed using mathematical approximations. In the image-based approach, image processing techniques were applied to isolate the dewatered region, and the relationship between the pixel count and actual filtrate volume was analyzed to estimate the cake moisture based on the calculated filtrate volume per minute. In the pressure-based approach, two models were proposed: one that directly estimates the cake moisture from the differential pressure, and another that models the relationship among the differential pressure, filtrate volume, and cake moisture. Unlike complex machine learning techniques, the proposed methods employ simple and interpretable mathematical functions, offering both practicality and reliability. Validation using real-world operational data confirmed the accuracy and effectiveness of the proposed approaches.