ИССЛЕДОВАНИЕ ОСОБЕННОСТЕЙ РАСПРЕДЕЛЕНИЯ ОШИБОК ПРИ МОДУЛЯРНЫХ ВЫЧИСЛЕНИЯХ В ПЕРСПЕКТИВНОЙ АСУ ТП НЕФТЕГАЗОВОГО КОМПЛЕКСА

Background The use of modular arithmetic as the mathematical basis for highly reliable calculations in the neural networks of promising automated process control systems (APCS) of the oil and gas industry is a well-known fact. The APCS with neural network organization provides fault-resistant architecture and its insensitivity to failures, i.e., failure of one or more elements of the architecture does not interfere with the task fulfillment and only slightly affects its quality. However, the study of the nature of error distribution by intervals within the range of the number system to the residue seems important for the development of the theory of corrective codes. Aims and Objectives To develop the theory of algorithmic design, and to study the principles of constructing fault-resistant computing structures based on neural networks in the basis of the modular arithmetic corrective codes that operate in the digital signal processing systems, in order to ensure failure safety of APCS in the oil and gas industry. Results On the basis of analyzing error distribution by intervals of the range of modular number system to the residue with a different number of control bases it was proved that the corrective abilities allow the neural network to locate and correct not only single, but also multiple distortions. The structure of detection and localization of errors is easy to implement on modern crystals of programmable logic with FPGA architecture of series PLIS Xilinx Virtex 7.