Dynamic measurement error correction on the basis of recurrent neural network

Abstract

Neural networks along with expert adaptive regulators and systems with associative memory form the basis of intelligent technologies for information management and processing [1], [2]. This paper introduces an algorithm for the dynamic measurement error correction based on a neural network (NN) inverse model of a measuring transducer (MT). It uses the reduction principle in the compensating filter for proper operation with high order MTs and noises filtering based on a NN model of an IIR filter. In this regard, the development of dynamic models of measuring systems based on recurrent NN and algorithms for dynamic measurement data processing using NN technologies is one of the promising directions for the development of the modern measuring equipment intellectualization. Successful solution to this task will significantly improve the metrological characteristics and efficiency of existing MTs without significant material costs due to deep mathematical processing of the measurement results.