Search academic texts

Information × Registration Number 0224U033600, (0124U004932) , R & D reports Title Development of a methodology for calculating the microkinetic models of chemical synthesis using physics-guided machine learning methods: third stage popup.stage_title Розвинення методики "неявного" диференціювання для розрахунку значень похідних. Розробка та тренування додаткових нейромереж для окремих доданків рівняння. Head Druchok Maksym Yu., Кандидат фізико-математичних наук Registration Date 31-12-2024 Organization Institute of Condensed Matter Physics of the National Academy of Sciences of Ukraine popup.description1 Research on approaches to calculating derivatives for a function approximated by a neural network. High-precision calculation of the fraction of vacant catalytic centers and its derivatives on thermodynamic and catalytic parameters in a microkinetic model of Fischer-Tropsch synthesis using physics-informed neural networks. popup.description2 This work is a continuation of the previous two projects, where it was proposed to use a neural network as an approximation function for calculating the microkinetic model of Fischer-Tropsch synthesis within a physics-informed neural network approach. An equation with rearranged parameters was proposed and a network was trained as its approximation. The results of the network were in good agreement with classical root-finding methods, however, the values of the derivatives calculated from the neural network contained large errors. In this work, it was shown that the use of the implicit differentiation approach allows to significantly increase the accuracy of calculating the derivatives, while not significantly increasing the calculation time.  Product Description popup.authors Demchuk Taras V. popup.nrat_date 2024-12-31 Close