Using Machine Learning in Electrical Tomography for Building Energy Efficiency through Moisture Detection

Wet foundations and walls of buildings significantly increase the energy consumption buildings, drying is one priority activities as part thermal modernization, along with insulation facades. This article discusses research findings detecting moisture decomposition within building utilizing electrical impedance tomography (EIT) deep learning techniques. In particular, focus was on algorithmic models whose task transforming voltage measurements into spatial EIT images. Two homogeneous networks were used: CNN (Convolutional Neural Network) LSTM (Long-Short Term Memory). addition, a new heterogeneous (hybrid) network built layers. Based reference reconstructions’ simulation data, three separate neural models: CNN, LSTM, hybrid model (CNN+LSTM), trained. Then, based popular measures such mean square error or correlation coefficient, quality assessed The obtained results showed that have great potential for solving tomographic inverse problem. Furthermore, it has been proven proper joining layers can improve effect reconstructions.