Hybrid machine learning in electrical impedance tomography

Artificial intelligence plays an increasingly important role in industrial tomography. In industry, various types of tomography can be used, where one of the criteria for classification may be a physical phenomenon. Thus, it is possible to distinguish computed tomography, impedance tomography, ultrasound tomography, capacitance tomography, radio-tomographic imaging, and others. The research described in this paper focuses on the EIT method used to imaging reactors' interior and industrial vessels. Inside the tested reactor, there may be a liquid of various densities containing solid inclusions or gas bubbles. The presented research presents the concept of transforming measurements into tomographic images using many known, homogeneous methods simultaneously. It is assumed that there is no single method of solving the inverse problem for all possible measurement cases. Depending on the specifics of the studied case, various methods generate reconstructions that differ in terms of accuracy and resolution. The presented research proves that the proposed approach justifies the increase in computational complexity, ensuring higher quality of tomographic images.