The use of the multi-sequential LSTM in electrical tomography for masonry wall moisture detection

Abstract

The study describes an innovative approach to reconstructing moisture distribution within walls using EIT, extending beyond the conventional single measurement vector methods. The research introduces a unique strategy involving a series of five measurement vectors, each deliberately disrupted by noise, to mirror real-life conditions more accurately. The aim was to compare the efficacy of this sequential approach with traditional single vector techniques in terms of accuracy and robustness in mapping moisture in masonry structures. The effectiveness of these reconstructions was evaluated based on various quality indicators. A multi-branch LSTM network was employed to model the moisture distribution reconstruction. The validation process included a comparative analysis using results from dielectric and gravimetric methods, enhanced with thermal imaging. The findings indicate that the multi-sequential reconstructions provide more detailed and accurate EIT representations than single-vector methods, demonstrating the increased accuracy and reliability of EIT reconstructions using this innovative technique.