Damp walls are a significant problem that affects not only historical buildings. The effects of moisture inside walls include premature degradation of the structure and paintwork as well as health hazards for people staying inside the rooms (fungi, microorganisms, allergens). To effectively remove moisture, it is necessary to identify the areas where it occurs. Tomography is the only nondestructive method that allows for imaging the interior of walls. It is not common due to the low image resolution [1]. The aim of the research presented is to present a new concept of impedance tomography, taking into account many measurement sequences at different frequencies of electric current. A neural network with LSTM (Long Short-Term Memory) layers was used to transform the measurements into images. A comparison of the results of the new approach proves the advantage of the multi-frequency method over the traditional method, which brings closer the breakthrough moment in the dissemination of tomography as the main method of imaging moisture in walls.

The study investigates the effectiveness of a differential neural network architecture in ultrasonic tomography (UST) for industrial applications. The proposed model employs a dual-branch structure, where each branch independently processes identical input data before passing the outputs to a differential layer. This approach enhances the model's ability to capture residual components, improving the reconstruction of tomographic images. Experiments were conducted using a tomographic system with 16 transducers, generating training and validation datasets. Comparative analysis between a differential LSTM-based network and a standard LSTM model demonstrated that the differential architecture achieved superior reconstruction quality. The results confirm that this approach enhances accuracy, sharpness, and overall image clarity, making it a promising solution for improving UST image reconstruction.

This study presents a novel technique for reconstructing the internal structures of industrial tank reactors using electrical impedance tomography (EIT). The method uses three different measurement vectors, each corresponding to different electrical frequencies---100 kHz, 50 kHz, and 10 kHz---to improve the accuracy and reliability of EIT reconstructions. The goal was to get the most out of both the resistive and reactive data from the EIT system by using machine learning methods that took frequency-specific data into account. This data was shown as complex numbers. To process the multi-frequency data collected from the measurements, an LSTM network was used. The results show that the multi-frequency model significantly outperforms single-frequency methods in terms of reconstruction accuracy.

This study aims to improve the resolution of reconstructed images from industrial ultrasonic tomography (UST) by determining the most effective neural network structure for solving the inverse problem based on the measurements taken. The study analyzed three types of neural networks: Artificial Neural Networks, Convolutional Neural Networks (CNNs) and a hybrid of CNNs and Long Short-Term Memory networks (LSTM). After evaluating the reconstructions and quality indicators, the CNN-LSTM combination provided the most accurate image reconstructions of the industrial ultrasound tomography, highlighting the importance of selecting an appropriate neural network to improve the resolution of the reconstructed images.