Using artificial neural networks to reduce nonlinearity of measuring devices

Abstract

The article discusses methods for reducing the impact of nonlinearity in the transformation function of measurement devices on the accuracy of measurement results by applying an additional correction device that implements a dependency that is inverse to the transformation function. The aim of the research is to explore the possibilities of using artificial neural networks, specifically multilayer perceptrons and radial basis function networks, as such correctors. The effectiveness of the proposed correction methods for the transformation function has been investigated through simulation computer modeling, examining the impact of the type of nonlinearity on the quality of such correction. A comparative analysis was carried out with traditional approaches, specifically a corrector based on polynomial approximation. The simulation results indicate that the accuracy of neural network correctors is comparable to that of polynomial correctors, and in some cases, even superior. This opens up prospects for a broader application of such modern measurement data processing methods as artificial neural networks in measurement technology.